Download Understanding Farmers` Perceptions and Adaptations to

IFPRI Discussion Paper 00849
February 2009
Understanding Farmers' Perceptions and Adaptations to
Climate Change and Variability
The Case of the Limpopo Basin, South Africa
Glwadys Aymone Gbetibouo
Environment and Production Technology Division
INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE
The International Food Policy Research Institute (IFPRI) was established in 1975. IFPRI is one of 15
agricultural research centers that receive principal funding from governments, private foundations, and
international and regional organizations, most of which are members of the Consultative Group on
International Agricultural Research (CGIAR).
FINANCIAL CONTRIBUTORS AND PARTNERS
IFPRI’s research, capacity strengthening, and communications work is made possible by its financial
contributors and partners. IFPRI receives its principal funding from governments, private foundations,
and international and regional organizations, most of which are members of the Consultative Group on
International Agricultural Research (CGIAR). IFPRI gratefully acknowledges the generous unrestricted
funding from Australia, Canada, China, Finland, France, Germany, India, Ireland, Italy, Japan,
Netherlands, Norway, South Africa, Sweden, Switzerland, United Kingdom, United States, and World
Bank.
AUTHOR
Glwadys Aymone Gbetibouo, Centre for Environmental Economics and Policy in Africa
Department of Agricultural Economics, University of Pretoria
[email protected] or [email protected]
Notices
1
Effective January 2007, the Discussion Paper series within each division and the Director General’s Office of IFPRI
were merged into one IFPRI–wide Discussion Paper series. The new series begins with number 00689, reflecting the
prior publication of 688 discussion papers within the dispersed series. The earlier series are available on IFPRI’s
website at www.ifpri.org/pubs/otherpubs.htm#dp.
2
IFPRI Discussion Papers contain preliminary material and research results. They have not been subject to formal
external reviews managed by IFPRI’s Publications Review Committee but have been reviewed by at least one
internal and/or external reviewer. They are circulated in order to stimulate discussion and critical comment.
Copyright 2009 International Food Policy Research Institute. All rights reserved. Sections of this material may be reproduced for
personal and not-for-profit use without the express written permission of but with acknowledgment to IFPRI. To reproduce the
material contained herein for profit or commercial use requires express written permission. To obtain permission, contact the
Communications Division at [email protected]
Contents
Acknowledgments
v Abstract
vi Abbreviations and Acronyms
vii 1. Introduction
1 2. Adaptation to Climate Change: A Theoretical Perspective
2 3. The Study Area: Limpopo River Basin in South Africa
4 4. The Data
6 5. Assessing Farmers’ Perceptions to Climate Change and Variability
7 6. Modeling Farmers’ Adaptation Options to Climate Change and Variability
15 7. Conclusions and Policy Implications
28 Appendix A: Questionnaire on Climate Change and Adaptation Options
29 Appendix B: Supplementary Tables
30 References
37 iii
List of Tables
1.
Limpopo Basin areas
4 2.
Perception of changes in temperature at provincial level (%)
8 3.
Analysis of temperature data from 1960 to 2003
9 4.
Perceptions of changes in rainfall (%)
10 5.
Analysis of the rainfall data from 1960 to 2003
11 6.
Moran’s I Test for spatial correlation of climate change perception
13 7.
Results of the seemingly unrelated biprobit model of farmers' perception of change in the
climate, Limpopo River Basin
14 8.
Barriers to adaptation in the Limpopo River Basin (% of the respondents)
15 9.
Adaptations options in response to change in temperature (% of respondents)
16 10. Adaptations in Response to Changes in Rainfall (% of respondents)
16 11. Variables hypothesized to affect adaptation decisions by farmers in the Limpopo River Basin
22 12. Results of the Heckman probit model of adaptations behavior in the Limpopo River Basin
24 13. Results marginal effects of the MNL adaptation model, Limpopo River Basin
27 B.1. Perceptions of changes in temperature by farmer experience (%)
30 B.2. Perceptions of changes in rainfall by farmer experience (%)
30 B.3. Perceptions of changes in temperature by farmer education level (%)
30 B.4. Correlation matrix of the independent variables of the adaptation model
31 B.5. Summary statistics of the variables for the adaptation model
33 B.6. Results of the multinomial logit adaptation model, Limpopo River Basin
35
List of Figures
1.
Provinces and WMAs in the Limpopo River Basin area
5 2.
Farmers’ perceptions of changes in temperature in the Limpopo River Basin
7 3.
Trend of temperature data for the Limpopo River Basin: 1960–2003
8 4.
Farmers' perceptions of changes in precipitation in the Limpopo River Basin
10 5.
Limpopo River Basin rainfall trend (1960-2003)
12 iv
ACKNOWLEDGMENTS
This work is supported by the Federal Ministry for Economic Cooperation and Development, Germany,
under the project Food and Water Security under Global Change: Developing Adaptive Capacity with a
Focus on Rural Africa, which forms part of the CGIAR Challenge Program on Water and Food. Support
has also been received under the IFPRI DGO Small Grants Initiative Program. The author would like to
thank Kato Edwards and Claudia Ringler for their helpful comments and directions on earlier versions of
the paper. Special thanks go to Elizabeth Bryan and Wisdom Akpalu for their assistance on numerous
technical points.
v
ABSTRACT
Climate change is expected to have serious environmental, economic, and social impacts on South Africa.
In particular, rural farmers, whose livelihoods depend on the use of natural resources, are likely to bear
the brunt of adverse impacts. The extent to which these impacts are felt depends in large part on the extent
of adaptation in response to climate change. This research uses a “bottom-up” approach, which seeks to
gain insights from the farmers themselves based on a farm household survey. Farm-level data were
collected from 794 households in the Limpopo River Basin of South Africa for the farming season 2004–
2005. The study examines how farmer perceptions correspond with climate data recorded at
meteorological stations in the Limpopo River Basin and analyzes farmers’ adaptation responses to climate
change and variability. A Heckman probit model and a multinomial logit (MNL) model are used to
examine the determinants of adaptation to climate change and variability. The statistical analysis of the
climate data shows that temperature has increased over the years. Rainfall is characterized by large
interannual variability, with the previous three years being very dry. Indeed, the analysis shows that
farmers’ perceptions of climate change are in line with the climatic data records. However, only
approximately half of the farmers have adjusted their farming practices to account for the impacts of
climate change. Lack of access to credit was cited by respondents as the main factor inhibiting adaptation.
The results of the multinomial logit and Heckman probit models highlighted that household size, farming
experience, wealth, access to credit, access to water, tenure rights, off-farm activities, and access to
extension are the main factors that enhance adaptive capacity. Thus, the government should design
policies aimed at improving these factors.
Keywords: climate change and variability, perception, adaptation, agriculture
vi
ABBREVIATIONS AND ACRONYMS
WMA
CEEPA
SAWS
MNL
MNP
GHK
IISHK
water management areas
Center for Environmental Economics and Policy in Africa
South African Weather Service
multinomial logit
multinomial probit
Geweke-Hajivassiliou-Keane
irrelevant alternatives
vii
1. INTRODUCTION
Adaptation is widely recognized as a vital component of any policy response to climate change. Studies
show that without adaptation, climate change is generally detrimental to the agriculture sector; but with
adaptation, vulnerability can largely be reduced (Easterling et al. 1993; Rosenzweig and Parry 1994;
Smith 1996; Mendelsohn 1998; Reilly and Schimmelpfennig 1999; Smit and Skinner, 2002). The degree
to which an agricultural system is affected by climate change depends on its adaptive capacity. Indeed,
adaptive capacity is the ability of a system to adjust to climate change (including climate variability and
extremes) to moderate potential damage, to take advantage of opportunities, or to cope with the
consequences (IPCC 2001). Thus, the adaptive capacity of a system or society describes its ability to
modify its characteristics or behavior so as to cope better with changes in external conditions.
Adaptation to climate change requires that farmers first notice that the climate has changed, and
then identify useful adaptations and implement them (Maddison 2006). Many agricultural adaptation
options have been suggested in the literature. They encompass a wide range of scales (local, regional,
global), actors (farmers, firms, government), and types: (a) micro-level options, such as crop
diversification and altering the timing of operations; (b) market responses, such as income diversification
and credit schemes; (c) institutional changes, mainly government responses, such as removal-preserve
subsidies and improvement in agricultural markets; and (d) technological developments—the
development and promotion of new crop varieties and advances in water management techniques (Smith
and Lenhart 1996; Mendelsohn 2001; Smit and Skinner 2002; Kurukulasuriya and Rosenthal 2003). Most
of these represent possible or potential adaptation measures rather than ones actually adopted. Indeed,
there is no evidence that these adaptation options are feasible, realistic, or even likely to occur.
Furthermore, they would only be possible with complete and accurate knowledge of future climatic
conditions, which is why these were aptly named “clairvoyant farmer” scenarios (Risbey et al. 1999, cited
by Belliveau et al. 2006). Thus, climate change impact studies often assume certain adaptations and little
explicit examination of how, when, why, and under what conditions adaptation actually occurs in
economic and social systems.
The present research, as part of a more recent strand of adaptation research, seeks to investigate
actual adaptations at the farm level, as well as the factors that appear to be driving them (e.g. Smit et al.
1996; Brklacich et al. 1997; Belliveau et al. 2006; Maddison 2006).
Based on the case of farmers in the Limpopo River Basin in South Africa, this paper intends to
capture the extent of farmers’ awareness and perceptions of climate variability and change and the types
of adjustments they have made in their farming practices in response to these changes. The analyses in
this study are based on a farm household survey of a total of 794 farmers conducted between August and
November 2005 in the Limpopo River Basin of South Africa.
The remainder of the paper is structured as follows. The next section gives brief theoretical
insights on research of adaptation to climate change. In Section 3, we present the study area, and in
Section 4 the data used in the study are discussed. Section 5 presents the assessment of farmers’
perceptions of climate change and variability. Section 6 presents the analytical and empirical adaptation
model. Finally, Section 7 concludes and outlines the policy implications of the study.
The primary hypothesis is that farmers adapt to perceived climate change and variability. The
analysis is conducted in two stages. First, it is determined whether the climate has changed, whether the
farmers perceive climate change and variability, and what characteristics differentiate farmers who
perceived changes from those who did not. Second, the determinants of adaptation are examined. Not all
of the farmers who perceived climate change will respond by taking adaptation measures. Here it is
argued that farmers who perceived climate changes and responded share some common characteristics.
Therefore, there is a need to understand the reasons underlying their response (or failure to respond for
those who did not adapt). Furthermore, adaptation to climate change requires farmers to choose from
among a set of adaptation options (practices and technologies) available in their region. By identifying the
important determinants of choosing any of the adaptation options, this paper provides important policy
information on how to promote various adaptations to climate change in rural South Africa.
1
2. ADAPTATION TO CLIMATE CHANGE: A THEORETICAL PERSPECTIVE1
2.1. Research Methods
Research on climate change–agriculture interactions has evolved from a “top-down” approach to a
“bottom-up” approach. The top-down mode starts with climate change scenarios, and estimates impacts
through scenario analysis, based on which possible adaptation practices are identified. The bottom-up
approach, on the other hand, takes on a vulnerability perspective where adaptation strategies are
considered more as a process involving the socioeconomic and policy environments, producers’
perceptions, and elements of decision-making (Bryant et al. 2000; Wall and Smit 2005; Belliveau et al.
2006).
In the top-down, scenario-based approach, adaptations are assumed and are invariably treated as
primarily technical adjustments (for example, changing to different crops, adopting efficient irrigation
systems, or altering production systems) to the impacts identified. Most of these adaptations represent
possible or potential adaptation measures, rather than measures that have actually been adopted. Indeed,
there is no evidence that these adaptation options are feasible, realistic, or even likely to occur.
Furthermore, they would only be possible with complete and accurate knowledge of future climatic
conditions, which is why they have been aptly named “clairvoyant farmer” scenarios (Risbey et al. 1999,
cited by Belliveau et al. 2006). This approach can be found in spatial analysis, climate impact modeling,
and Ricardian studies. Most studies on climate impacts using the top-down approach carried out in South
Africa (Schulze et al. 1993; Erasmus et al. 2000; du Toit et al. 2001; Kiker 2002; Kiker et al. 2002;
Poonyth et al. 2002; Deressa 2003; Gbetibouo and Hassan 2005) predicted adverse impacts on the
agricultural sector with significant adverse effects on crop yields and marginal crop areas in the western
part of the country, which would become unsuitable for the production of maize, the main staple crop.
Vulnerability studies have shifted the focus of research from the estimation of impacts to the
understanding of farm-level adaptation and decision making. This research explores actual adaptation
behavior based on the analysis of farmer decisions in the face of variable conditions through survey data
analysis, in-depth interviews, and focus group discussions with farmers and other farms experts (Smit et
al. 1996; Brklacich et al. 1997; Chiotti et al. 1997; Maddison 2006; Belliveau et al. 2006). According to
Bryant et al. (2000), these studies have raised new research questions regarding how farmers perceive
climatic change and variability; have identified those climatic properties that are of most importance to
farmers in their decision making; and have suggested the types of adaptive responses that can be
anticipated.
2.2. Agricultural Adaptation to Climate Change
Agricultural change does not involve a simple linear relationship between changes in a farmer’s decisionmaking environment and farm-level change.
One important issue in agricultural adaptation to climate change is the manner in which farmers
update their expectations of the climate in response to unusual weather patterns. Referring to Kolstad et
al. (1999), Maddison (2006) discusses what he calls “the transitional cost” of adapting to climate change.
The transitional cost is the difference between the maximum value of net revenues per acre following
perfect adaptation and the net revenues actually experienced by farmers given that their expectations of
(and therefore response to) climate change lag behind actual climate change. A farmer may perceive
several hot summers but rationally attribute them to random variation in a stationary climate. One could
argue that farmers engage in simple Bayesian updating of their prior beliefs according to the standard
formula. If so, the process of updating is likely to be slow, and therefore one should not expect decades of
information to be thrown out overnight. However, there is evidence that farmers did not update their
priors in this way. Indeed, farmers place more weight on recent information than is efficient.2
1
2
This section draws heavily on Bryant et al. (2000), Belliveau et al. (2006), and Maddison (2006).
For reference see Smit et al. (1997).
2
Another important issue related to adaptation in agriculture pointed out by Bryant et al. (2000) is
how perceptions of climate change are translated into agricultural decisions. If farmers learn gradually
about the change in climate, Maddison (2006) argues that they will also learn gradually about the best
techniques and adaptation options available. According to him, farmers learn about the best adaptation
options through three ways: (1) learning by doing, (2) learning by copying, and (3) learning from
instruction. There is recognition that farmers’ responses vary when faced with the same stimuli. Such
varied responses, even within the same geographic area, are partly related to the variety of agricultural
systems involved and the different market systems in which farmers operate (Bryant et al. 2000). A more
important factor of varied farmers’ responses is the differences between farmers in terms of personal
managerial and entrepreneurial capacities and family circumstances. Also, farmers can be influenced by
their peers’ perceptions and by values present in their communities as well as their professional
associations. A review of literature on adoption of new technologies identified farm size, tenure status,
education, access to extension services, market access and credit availability, agroclimatic conditions,
topographical features, and the availability of water as the major determinants of the speed of adoption
(Maddison 2006).
This paper adopts the bottom-up approach that seeks to investigate actual adaptations at the farm
level, as well as the factors that appear to be driving them. Based on the case of farmers in the Limpopo
River Basin in South Africa, this paper intends to capture the extent of farmers’ awareness and
perceptions of climate variability and change, and the types of adjustments they have made in their
farming practices in response to these changes.
3
3. THE STUDY AREA: LIMPOPO RIVER BASIN IN SOUTH AFRICA
The Limpopo River is one of the major river systems in Southern Africa. Originating in South Africa’s
Witwatersrand region, the Limpopo River is about 1,700 kilometers long, drains an area of about 415,500
square kilometers, and is shared by Botswana, Mozambique, South Africa, and Zimbabwe. South Africa
occupies about 46 percent of the total area of the basin (Table 1).
Table 1. Limpopo Basin areas
Country
Total area Area of the Percentage Percentage Irrigation
Area
of the
of total of total area potential
country
under
country
area of
within the
of country (hectares) irrigation
(hectares)
(km2)
basin (km2) basin (%)
(%)
Average annual
rainfall in the basin
area
(mm)
min
max
mean
Botswana
581,730
80,118
19.9
13.8
5,000
1,381
290
555
425
Zimbabwe
390,760
51,467
12.38
13.2
10,900
200
300
635
465
Mozambique
801,590
84,981
21.1
10.6
148,000
40,000
355
865
535
185,298
46.31
15.2
131,500
198,000
290
1,040
590
401,864
100.0
--
295,400
241,381
290
1,040
530
South Africa 11,221,040
Total
12,995,120
Source: Food and Agriculture Organization of the United Nations,
www.fao.org/documents/show_cdr.asp?url_file=/docrep/W4347E/w4347e0p.htm
In South Africa, the Limpopo River Basin extends over four administrative provinces (Limpopo,
Mpumalanga, Gauteng, and North West) and five water management areas (WMAs).3 These are the (1)
Limpopo, (2) Luvuvhu/Letaba, (3) Crocodile (west) and Marico, (4) Olifants, and (5) Inkomati (Figure 1).
The climate in the basin ranges from temperate and semiarid in the south of the Limpopo WMA
and the east of the Crocodile (west) and Marico WMA to arid in the extreme north of the Limpopo WMA.
Mean annual precipitation ranges from 200 millimeters per year in the Limpopo WMA to more than
2,300 millimeters annually in the Luvuvhu/Letaba WMA, with temperatures ranging from 8 degrees
Celsius to more than 30 degrees Celsius in the northern parts of the Limpopo WMA. Among the five
WMAs, Inkomati has the highest mean annual runoff with 3,539 million cubic meters, followed by
Olifants (2,040 million), Luvuvhu/Letaba (1,185 million), Limpopo (986 million) and Crocodile (west)
and Marico (855 million).
South Africa and the Limpopo Basin feature both large commercial agricultural farms as well as
small-scale agriculture.
The diverse climate in the basin influences the vegetation and agricultural activities in the five
WMAs and the four provinces. Farming activities range from dry farming to intensive irrigation and
livestock production. For example, farming in the Limpopo WMA mainly focuses on livestock and
irrigation, while the Olifants WMA is favorable for dry-land farming and livestock as well as extensive
irrigation. Intensive irrigation is prevalent in the Crocodile (west) and Marico, where the irrigation sector
is the second largest water consumer in the WMA, using an estimated 33 percent of the total water use. In
the Limpopo WMA, crops grown include cotton, grain sorghum, and tobacco, as well as considerable
3
To facilitate the management of the scarce water resources, the country has been divided into 19 catchment-based water
management areas (WMAs). All except one WMA are interlinked with other areas through inter-catchment transfers. The
interlinking of catchments gives effect to one of the main principles of the country’s National Water Policy Acts, which
designates water as a national resource (DWAF 2004).
4
subsistence production. In the Luvuvhu/Letaba WMA, citrus and a variety of fruits plus commercial
forestry are prevalent. The Olifant WMA features trout and game farming.
Figure 1. Provinces and WMAs in the Limpopo River Basin area
Source: http://www.exittoafrica.com/media/images/maps/m-sa_provinces.gif and
http://www.dwaf.gov.za/Documents/Notices/Water%20Management%20areas%20engl.doc
5
4. THE DATA
4.1. Survey Data
The survey collected a large range of data. However, this study used principally the section of the survey
on perceptions of climate change, adaptations made by farmers, and barriers to adaptation. Open-ended
questions were used to ask farmers whether they had noticed long-term changes in mean temperature,
mean rainfall, the number of malaria cases, and vegetation cover over the past 20 years. Questions about
adaptation and the constraints to adaptation were also posed. The exact formulation of the questions is
included in Appendix A.
The empirical analyses of this paper used data obtained from an ongoing project entitled Food
and Water Security under Global Change: Developing Adaptive Capacity with a Focus on Rural Africa,
funded by the Advisory Service on Agricultural Research for Development of the German Government.
Under the project, a survey was carried out by the Center for Environmental Economics and Policy in
Africa (CEEPA), University of Pretoria, in collaboration with the International Food Policy Research
Institute (IFPRI) to analyze the potential impact of climate variability and climate change on household
vulnerability and farm production. The survey period was between August and November 2005 covering
the April/May 2004 to April/May 2005 agricultural season. In total, 794 surveys were completed in 19
districts of 4 provinces of South Africa. Farmers were carefully selected with the assistance of producers
associations and the National Department of Agriculture.
4.2. Meteorological Data
Monthly precipitation and temperature data was obtained from the South African Weather Service
(SAWS). The data covers the period from January 1960 to October 2003. To capture the provincial
temperature, a mean of all stations in each province was calculated. For the whole Limpopo River Basin
the same analysis was conducted.
6
5. ASSESSING FARMERS’ PERCEPTIONS TO CLIMATE CHANGE AND
VARIABILITY
5.1. Comparison between Perceptions of Changes in Climate and Meteorological
Stations’ Recorded Data
To assess farmers’ perceptions of climate change and variability, we first look at how climate data
recorded at meteorological stations evolved (trends and variability) and how farmers perceived these
changes. Tests were undertaken for linear trend in annual means and seasonal means of temperature, and
total annual and seasonal rainfall both at the Limpopo River Basin level and at the provincial level.
Descriptive statistics based on summary counts of the questionnaire structure are used to provide insights
into producers’ perceptions of climate change and variability. In the literature several studies have
undergone the same type of analysis.
For example, Vedwan and Rhoades (2001) examine how apple farmers in the western Himalayas
of India perceive climatic change. This is done by comparing the locally idealized traditional weather
cycle with climate change as perceived by the farmers of the region using snowfall and rainfall data to
measure the accuracy of perceptions. Hageback et al. (2005) assess small-scale farmers’ perceptions of
climate change in the Danagou watershed in China by comparing the local precipitation and temperature
data trend with the responses given by farmers to the question “Do you feel any changes in the weather
now compared to 20 years?” They conclude that farmers’ perceptions of climatic variability correspond
with the climatic data records. Another study by Maddison (2006), using data for over 9,500 farmers from
eleven African countries, compared the probability that the climate has changed, as revealed by an
analysis of the statistical record, with the proportion of individuals who believe that such a change has in
fact occurred to assess farmers’ perceptions of climatic change.
5.1.1. Temperature Changes
About 95 percent of the farmers interviewed perceived long-term changes in temperature. Most of them
(91 percent or 686 farmers) perceive the temperature in the Limpopo Basin to be increasing. Only 1.5
percent noticed the contrary, a decrease in temperature. Across the four provinces of the basin, the same
pattern is noticed with the exception of Gauteng, where 16 percent, which represents 7 farmers
interviewed, have not noticed any changes in the temperature (Figure 2 and Table 2).
Figure 2. Farmers’ perceptions of changes in temperature in the Limpopo River Basin
% of respondents
100
90
Increased
80
70
Decreased
60
50
40
Altered Climatic Range/
More or Less Extreme
Others
30
20
No change
10
0
Don't Know
Perception Temperature
7
Table 2. Perception of changes in temperature at provincial level (%)
Total Basin
Limpopo
Gauteng
North West
Mpumalanga
Increased
89.32
90.67
66.67
89.52
91.52
Decreased
1.56
1.2
4.44
2.86
1
More or less Extreme
1.56
1.67
4.44
0
1.5
Others
2.21
1.44
8.89
4.76
1
No change
5.08
4.78
15.56
2.89
4.5
Don't know
0.26
0.24
0
0
0.5
The statistical record of temperature data from the Limpopo River Basin between 1960 and 2003
shows an increasing trend, with the increase mostly in the summer. In 43 years, the temperature has risen
around 1 degree Celsius. An analysis at the provincial level shows the same general trend of increasing
temperature (Figure 3). However, in the North West province the trend is not significant. Mpumalanga
province had a higher increase in temperature of around 1.83 degrees Celsius. In Limpopo, Gauteng, and
Mpumalanga provinces, the increased trend is occurring mostly during the winter season (Table 3).
Thus, farmers’ perceptions appear to be in accordance with the statistical record in the region.
Figure 3. Trend of temperature data for the Limpopo River Basin: 1960–2003
30
y = 0.0201x - 17.488
R2 = 0.0873
temperature degree celsius
25
20
Annual average temperature
y = 0.021x - 23.37
R2 = 0.2572
Summer Average
Temperature
15
y = 0.0271x - 38.719
R2 = 0.3291
Winter Average Temperature
10
5
0
1950
1960
1970
1980
1990
2000
year
8
2010
Table 3. Analysis of temperature data from 1960 to 2003
Temperature data for the total Limpopo River Basin
Temperature
Yearly
Summer
Mean (0C)
18.15
22.31
0.52
0.85
Standard deviation (0C)
17.15
20.6
Minimum temperature (0C)
19.43
24.09
Maximum temperature (0C)
0.02*
0.02*
Trend (0C/year or season)
Correlation
0.51
0.30
0.94
2.04
Total change calculated from the trend
(0C/43 years)
Temperature data for the Limpopo province
Temperature
Yearly
Summer
Mean (0C )
20.83
23.93
0.57
0.64
Standard deviation (0C )
19.18
22.67
Minimum temperature (0C)
22.18
25.35
Maximum temperature (0C)
0.02*
0.01*
Trend (0C/year or season)
Correlation
0.47
0.29
0.83
0.76
Total change calculated from the trend
(0C/43 years)
Temperature data for the North West province
Temperature
Yearly
Summer
Mean (0C )
18.39
22.31
0.48
0.68
Standard deviation (0C )
17.28
20.98
Minimum temperature (0C)
19.51
23.86
Maximum temperature (0C)
0.009
0.007
Trend (0C/year or season)
Correlation
0.24
0.14
0.28
0.85
Total change calculated from the trend
(0C/43 years)
Temperature data for the Gauteng province
Temperature
Yearly
Summer
Mean (0C )
9.91
14.15
0.48
0.46
Standard deviation (0C )
8.56
12.11
Minimum temperature (0C)
10.75
15.09
Maximum temperature (0C)
0.02*
0.01*
Trend (0C/year or season)
Correlation
0.58
0.38
1.03
0.31
Total change calculated from the trend
(0C/43 years)
Temperature data for the Mpumalanga province
Temperature
Yearly
Summer
Mean (0C )
17.52
20.59
0.77
0.80
Standard deviation (0C )
16.08
18.94
Minimum temperature (0C)
19.17
22.15
Maximum temperature (0C)
0.04*
0.03*
Trend (0C/year or season)
Correlation
0.63
0.57
1.83
1.80
Total change calculated from the trend
(0C/43 years)
Notes: *P <0.05 Student’s t-test, N=43.
Total change is the difference between the trend line value of the first and last year.
9
Winter
15.04
0.59
13.93
16.41
0.03*
0.60
1.15
Winter
17.75
0.68
16.58
19.15
0.03*
0.55
1.02
Winter
14.49
0.47
13.64
15.45
0.01*
0.32
0.41
Winter
5.71
0.63
4.53
6.85
0.03*
0.58
1.68
Winter
13.90
0.91
12.22
15.66
0.05*
0.65
2.26
5.1.2. Precipitation Changes
In total, 97 percent of the respondents observed changes in rainfall patterns over the past 20 years, and 81
percent (or 624) noticed a decrease in the amount of rainfall or a shorter rainy season. Almost 5 percent of
the informants noticed a change not in the total amount of rainfall but in the timing of the rains, with rains
coming either earlier or later than expected. The same pattern is observed across the four provinces with
slight differences (Figure 4 and Table 4). A change in the timing of rainfall was mentioned by 7 percent
of the farmers in Mpumalanga, 12 percent in North West, and 24 percent in Gauteng. Many respondents
observed that the main rainfall season, which is the summer, is coming late and is also shorter.
Figure 4. Farmers' perceptions of changes in precipitation in the Limpopo River Basin
Increased
Decreased
2%
0%
Change in the timing of
rains/earlier/later/erratic
1%
5%
1%
7%
2%
Change in frequency of
droughts/floods
Others
No change
82%
Don't know
Decreased rainfall and
change in the timing
Table 4. Perceptions of changes in rainfall (%)
Increased
Total Basin
Limpopo
Gauteng
North West
Mpumalanga
1.69
1.68
2.22
0.94
2
Decreased
81.25
91.61
57.78
62.26
75
Change
in timing
of rains
(earlier/
later/
erratic)
5.34
0.72
24.44
12.26
7
Change in
frequency
of
droughts/
floods
0.65
0.72
0
1.89
0
Others
1.43
0.96
2.22
3.77
1
No
change
2.21
2.4
4.44
2.83
1
Don't
know
Decrease
rainfall
and
change in
the timing
0.39
0.24
0
0.94
0.5
7.03
1.68
8.89
15.09
13.5
The recorded data on rainfall from 1960 to 2003 shows that about 85 percent of the rainfall
occurs during summer months (October to March). Also, there is no statistically significant trend in the
data. The exception is during the winter season, when data show a decreasing trend. The correlation
between rainfall and time is also insignificant. Indeed, there is a large variability in the amount of
precipitation from year to year. The same pattern is observed in each province (Table 5 and Figure 5).
The high proportion of farmers noticing a decrease in precipitation could be explained by the fact
that during the last few years of the study (2001 to 2003), there was a substantial decrease in the amount
of rainfall (Figure 5). Thus, farmers’ perceptions of a reduction in rainfall over the 20-year period is
10
explained by the fact that, as Maddison (2006) noticed, some farmers place more weight on recent
information than is efficient.
Table 5. Analysis of the rainfall data from 1960 to 2003
Rainfall data for the Limpopo River Basin
Rainfall
Yearly
Summer
Mean (mm)
681.76
579.89
Percentage of yearly total
85
Standard deviation (mm)
123.53
128.97
Minimum rainfall (mm)
425.858
352.97
Maximum rainfall (mm)
967.09
906.25
Trend (mm/year or season)
0.38
0.7
Correlation
0.04
0.07
Total change calculated from the trend
-242.84
-255.87
(mm/43 years)
Total change calculated from the trend (%)
-37.64
-38.77
Rainfall data for the Limpopo province
Rainfall
Yearly
Summer
Mean (mm)
612.39
525.00
Percentage of yearly total
85.78
Standard deviation (mm)
143.15
153.44
Minimum rainfall (mm)
364.64
271.48
Maximum rainfall (mm)
1000.36
900.54
Trend (mm/year or season)
1.92
1.61
Correlation
0.17
0.13
Total change calculated from the trend
-304.59
-308.35
(mm/43 years)
Total change calculated from the trend (%)
-45.26
-47.23
Rainfall data for the North West province
Rainfall
Yearly
Summer
Mean (mm)
605.83
513.80
Percentage of yearly total
84. 8
Standard deviation (mm)
133.61
133.10
Minimum rainfall (mm)
341.85
320.96
Maximum rainfall (mm)
922.01
879.78
Trend (mm/year or season)
-1.09
-0.4
Correlation
-0.10
-0.03
Total change calculated from the trend
-272.86
-242.69
(mm/43 years)
Total change calculated from the trend (%)
-39.51
-39.88
Rainfall data for Gauteng province
Rainfall
Yearly
Summer
Mean (mm)
687.71
588.58
Percentage of yearly total
85.58
Standard deviation (mm)
133.91
135.72
Minimum rainfall (mm)
429.36
353.65
Maximum rainfall (mm)
971.45
965.29
Trend (mm/year or season)
0.33
0.85
Correlation
0.03
0.08
Total change calculated from the trend
-190.62
-162.17
(mm/43 years)
Total change calculated from the trend (%)
-28.26
-26.73
11
Winter
99.36
15
38.16
30.65
181.94
-0.91*
-0.27
-20.97
-18.79
Winter
85.64
14.22
33.108
26.50
159.56
-0.37
-0.09
-23.22
-22.56
Winter
89.95
15.2
52.11
17.30
224.006
-1.25*
-0.26
0.93
1.07
Winter
97.67
14.42
43.37
18.43
212.59
-1.11*
-0.27
-8.37
-8.55
Table 5. Continued
Rainfall data for Mpumalanga province
Rainfall
Yearly
Summer
Mean (mm)
821.11
692.16
Percentage of yearly total
84.29
Standard deviation (mm)
140.03
137.95
Minimum rainfall (mm)
554.93
442.57
Maximum rainfall (mm)
1197.05
1077.90
Trend (mm/year or season)
0.37
0.7
Correlation
0.03
0.07
Total change calculated from the trend
-361.70
-310.26
(0C/43 years)
Total change calculated from the trend (%)
-37.56
-40.22
Winter
124.19
15.71
44.46
41.20
255.27
-1.25*
-0.22
-53.25
-33.70
Notes: *P <0.05 Student’s t-test, N=43.
Total change is the difference between the trend line value of the first and last years.
Figure 5. Limpopo River Basin rainfall trend (1960-2003)
1200
rainfall (mm)
1000
y = 0.3881x + 673.23
R2 = 0.0016
800
Annual rainfall
600
Summer rainfall
Winter rainfall
400
y = 0.7073x + 564.33
R2 = 0.0047
y = -0.766x + 117.7
R2 = 0.0664
200
19
60
19
63
19
66
19
69
19
72
19
75
19
78
19
81
19
84
19
87
19
90
19
93
19
96
19
99
20
02
0
year
5.2. Spatial Clustering of Climate Change Perceptions
As shown in the above section, a large number of farmers believe the climate has become hotter and drier.
As suggested by Maddison (2006), this perception might be a case of prominence bias in questionnaires
dealing with climate change. It’s likely that some respondents provided answers during the interview that
the enumerators were more interested in hearing. Thus, validation of the respondent’s assessment of
climate change with his/her neighbors’ responses would provide more confidence that the responses were
objective and not subjective.
We employed Moran’s I4test for spatial autocorrelation with an inverse distance weights matrix
on the portion of farmers who perceive particular types of climate change within a given area. The results
(see Table 6) suggested that neighboring farmers agree that temperature is increasing and rainfall is
4
For further details on Moran I test see Cliff, A. D., Ord, J. K. 1981. Spatial processes, Pion.
Ho: spatial independence.
12
decreasing with a change in the timing. These results are evidence that farmers are capable of perceiving
changes in climate. Thus, neighboring farmers tell a consistent story.
Table 6. Moran’s I Test for spatial correlation of climate change perception
Perception of
temperature
Increased temperature
Decreased temperature
More or less extreme
No change
Moran I statistics
Perception of rainfall
Moran I statistics
0.044**
0.002
0.001
-0.003
Increased rainfall
Decreased rainfall
Change in the timing
Change in frequency of
droughts/floods
No change
-0.013
0.125**
0.051**
-0.007
0.003
Note: ** significant at 1% level * significant at 5% level
5.3. What Types of Farmers Perceive Climate Change?
To answer the questions regarding which types of farmers perceive climate change, farmers’ perceptions
of climate change have been classified according to their years of experience and their level of education.
In Appendix B, we distinguish the responses of farmers having less than 10 years, between 10 and 30
years, and 30 or more years of experience. For the level of education, we distinguish four classes: (1) no
formal education, (2) standard education, (3) secondary education, and (4) tertiary education. Using the
Kruskal-Wallis test, a nonparametric test, we assess if the perceptions of climate according experience
level and education level differed significantly.
The results show that a slightly higher proportion of farmers with more than 30 years of
experience claimed that temperature is increasing and rainfall is decreasing, and noted change in the
frequency of droughts and floods. Farmers with more than 30 years of experience are also less likely to
claim no change in temperature and no change in rainfall. However, the Kruskal-Wallis test indicated that
the views between experienced and inexperienced farmers are statistically not significant. The results also
indicated that there is statistically no difference between the views of the educated and less-educated
farmers.
The above results do not indicate whether the results are sensitive to other factors; therefore, we
analyze which types of farmers are likely to notice climate change (temperature and/or rainfall changes)
by running a probit model. Rather than present results for each category of climate change perception, we
limit the analysis to explaining the twin perception of change in both temperature and precipitation.
Because these two perceptions are likely to be correlated, we run a seemingly unrelated biprobit model.
The independent variables are education, farming experience, farm size, whether or not a crop farm, soil
characteristics, irrigation, access to extension services, access to climate information, and region dummy
for Gauteng. These results are adjusted for clustering at the district level on the assumption that the
responses from farmers in the same district are likely to be related anyway.
The results displayed in Table 7 below show the following:
1. Education seems to decrease the probability that the farmer will perceive long-term changes
in rainfall. Thus, educated farmers are more likely to see that rainfall does not have a
significant trend over the long run.
2. With experience, farmers are more likely to perceive change in temperature.
3. Farmers who have access to water for irrigation purposes are unlikely to perceive any change
in the climate whether in temperature or rainfall. Indeed, having access to water increases the
resilience of farmers to climate variability.
4. Access to extension, on the other hand, increases the probability of perceiving change in
temperature.
13
5. The results also confirm that being in Gauteng (the biggest province where agriculture is a
small part of the economy) decreases the probability of perceiving climate change.
6. Farmers with highly fertile soil are less likely to perceive change in temperature but more
likely to perceive change in rainfall.
Table 7. Results of the seemingly unrelated biprobit model of farmers' perception of change in the
climate, Limpopo River Basin
Education
Farming experience
Farm size
Crop farm
Infertile soil
Highly fertile soil
Access to water for irrigation
Access to extension services
Access to climate information
Gauteng dummy
Intercept
Log likelihood: -186.0339
Number of observations: 632
Athrho: 0.8027***
Rho: 0.6655
Perceive change in
temperature
-0.0049
0.0136*
0.2900
0.0822
-0.3838
-0.3231**
-0.5917**
0.3361**
-0.0101
-0.6374***
1.91923 ***
Perceive change in rainfall
-0.0371***
0.0048
-0.3474
-0.0219
0.0994
0.6542**
-0.7279**
0.2271
0.2044
.245423
2.4828***
Notes: The coefficient indicated the impact of a marginal change on the probability, while dF/dx is for discrete change of dummy
variable from 0 to 1.
Dummy variable for Gauteng is included because in the descriptive analysis, 16% of farmers from Gauteng did not perceive
climate change.
Clustering at district level.
Wald test of rho=0:
chi2(1) = 28.5094 Prob > chi2 = 0.0000
*** significant at 1% level; ** significant at 5% level; * significant at 10% level.
14
6. MODELING FARMERS’ ADAPTATION OPTIONS TO CLIMATE CHANGE AND
VARIABILITY
This section describes the approach adopted by the study to model adaptation options to climate change
and variability by farmers in the Limpopo River Basin as follows: (6.1) the adaptation choices in the
study area, (6.2) the analytical framework, (6.3) the factors hypothesized to influence farmers’ adaptation
options, and (6.4) the empirical models and the results.
Nevertheless, empirical assessment of actual adaptive behavior is advocated, even though such
behavior is place- and time-specific and more likely represents a response to interperiodic climatic
variability, as well as to multiple nonclimatic risks and opportunities (Belliveau et al. 2006).
Understanding the likely adaptive responses of farmers to anticipated climate change represents
serious challenges for researchers. One major challenge is to isolate the climate stimuli response from
other stimuli (market, policy, etc.) that farmers face in the real world. Secondly, farmers are more
concerned with and respond more to short-term climate variability than climate change. However, the
ability of farmers to cope with current climate variability is an important indicator of their capacity to
adapt to future climate change. Thirdly, as Belliveau et al. (2006) mention, a more fundamental barrier to
improved knowledge of climate change adaptation derives from the simple fact that humans can respond
in highly variable ways to similar external stimuli.
6.1. Adaptation Options in the Study Area
This section focuses on the various adjustments that farmers in the survey made in their farming activities
if they perceived changes in the climate. Even though a large number of farmers interviewed noticed
changes in climate, almost two-thirds did not undertake any remedial actions. More than 53 percent of
farmers cited lack of access to credit, poverty, and lack of savings as the main barriers to adaptation.
Despite perceiving a decrease in the volume of rainfall, 20 percent of farmers are not irrigating because
they do not have access to water. In the Limpopo province, the claim of no access to water is about 32
percent. Insecure property rights and lack of markets are also cited as significant barriers to adjustments.
Few farmers (1.9 percent) designated lack of information or knowledge of appropriate adaptation
measures as barriers to adaptations (Table 8).
Table 8. Barriers to adaptation in the Limpopo River Basin (% of the respondents)
Lack of
information
about longterm climate
change
Total Basin
6.03
Limpopo
4.32
North West
Lack of
knowledge
concerning
appropriate
adaptations
1.95
Lack of No access Insecure
property
credit or to water
rights
savings /
poverty
53.9
20.75
9.57
2.65
24.24
32.58
14.27
3.49
3.49
10.47
0.00
54.65
Gauteng
0.00
0.00
32
Mpumalanga
8.56
1.98
48.04
12
8.56
Lack of
Other No
barriers to
market access
adaptation
poor
transport links
6.21
10.99
0.78
7.97
8.31
1.16
9.3
22.09
0.00
4
20
10
5.92
1.32
10.3
13.10
23.03
Indeed, in the Limpopo River Basin only 30 percent of respondents made adjustments to their
farming practices in response to perceived increases in temperature, and 33 percent in response to changes
in rainfall patterns. A number of adaptation options are identified. However, the responses to perceived
rainfall and perceived temperature changes differ. The main adjustments in farming activities are
discussed below.
15
6.1.1. Farmers’ Responses in the Face of Increased Temperature
Eight adaptation measures could be identified in the Limpopo River Basin as farmers’ responses to
increased temperature (Table 9): planted different crops (6.86 percent), changed crop variety (3 percent),
changed planting dates (3.69 percent), increased irrigation (3.96 percent), used crop diversification (1
percent), changed the amount of land under cultivation or grazed (3.43 percent), and invested in livestock
by buying feed supplements (3.69 percent); other adaptation measures were cited by 5 percent of farmers.
Table 9. Adaptations options in response to change in temperature (% of respondents)
Change
crop
variety
Total Basin
Limpopo
North West
Gauteng
Mpumalanga
3.03
1.21
3.92
2.27
6.57
Irrigate
more
3.96
3.38
1.96
6.82
5.56
Feed
supplements
Different
crops
3.69
3.62
5.88
4.55
2.53
6.86
9.66
3.92
0.00
4.04
Crop
diversific
ation
(mixed/
multicropping)
0.53
0.97
0.00
0.00
0.00
Different
planting
dates
Change
amount of
land under
cultivation
or grazed
Other5
No
adaptation
3.69
3.62
0.98
6.82
4.55
3.43
4.11
1.96
2.27
3.03
5.01
4.83
2.94
6.82
6.06
69.39
67.87
78.43
70.45
67.68
6.1.2. Farmers’ Responses in the Face of Reduced Rainfall and Disrupted Rainfall Patterns
Among those who did adjust to changes in rainfall patterns (Table 10), 88 farmers (11.56 percent)
engaged in irrigation, 3.81 percent for a new water scheme, and 7.75 percent for increased in irrigation).
Farmers also used different crops (4.7 percent), shifted their planting dates to match the delay in rainfall
(3 percent), changed the amount of land under cultivation or grazed (2.76 percent), and invested in
livestock by buying feed supplements (2.23 percent).
The adaptations induced by perceptions of changing rainfall patterns seem to differ from those
induced by perceptions of changing temperature. While adopting a new crop variety is the main strategy
used to adapt to increasing temperature, building water-harvesting schemes is a popular adaptation
strategy to those experiencing the effects of decreased precipitation.
Table 10. Adaptations in Response to Changes in Rainfall (% of respondents)
Change Build a Irrigate Buy feed Different Different Change Other
No
crop
more supplements crops planting amount of
wateradaptation
variety harvesting
dates land under
scheme
cultivation
or grazed
0.66
3.81
7.75
2.23
4.99
4.73
2.76
5.12
Total Basin
67.94
0.72
3.61
4.82
2.41
6.75
3.13
4.34
4.34
Limpopo
69.88
0.00
1.94
13.99
3.88
2.91
3.88
0.00
4.85
North West
68.06
0.00
4.55
4.55
2.27
2.27
9.09
0.00
4.55
Gauteng
72.73
1.01
5.03
11.56
1.01
3.02
7.54
1.51
7.04
Mpumalanga
62.31
5
Other adaptation measures: (1) implement soil conservation techniques; (2) put trees for shading; (3) change from crops to
livestock; (4) reduce number of livestock; (5) migrate to urban area; and others.
16
6.2. Analytical Framework
The decision of whether or not to use any adaptation option could fall under the general framework of
utility and profit maximization.
Consider a rational farmer who seeks to maximize the present value of expected benefits of
production over a specified time horizon, and must choose among a set of J adaptation options.
The farmer i decides to use j adaptation option if the perceived benefit from option j is greater
than the utility from other options (say, k) depicted as
U ij  j X i   j   U ik  k X i   k , k  j ,
(1)
where Uij and Uik are the perceived utility by farmer i of adaptation options j and k, respectively; Xi is a
vector of explanatory variables that influence the choice of the adaptation option; βj and βk are
parameters to be estimated; and εj and εk are the error terms.
Under the revealed preference assumption that the farmer practices an adaptation option that
generates net benefits and does not practice an adaptation option otherwise, we can relate the observable
discrete choice of practice to the unobservable (latent) continuous net benefit variable as Yij = 1 if Uij> 0
and Yij = 0 if Uij< 0. In this formulation, Y is a dichotomous dependent variable taking the value of 1
when the farmer chooses an adaptation option in question and 0 otherwise.
The probability that farmer i will choose adaptation option j among the set of adaptation options
could be defined as follows:
P Y  1 / X   P U ij  U ik / X 
 P j X i   j   k X i   k  0 / X 
 P  j   k X i   j   k  0 / X 
(2)
 P(  * X i   *  0 / X ) F (  * X i ) ,
where ε* is a random disturbance term, β* is a vector of unknown parameters that can be interpreted as
the net influence of the vector of explanatory variables influencing adaptation, and F (β*Xi) is the
cumulative distribution of ε* evaluated at β*Xi. Depending on the assumed distribution that the random
term follows, several qualitative choice models such a linear probability, logit, or probit model could be
estimated (Greene 2003). The logit and probit models are the most common models used in the literature.
Indeed, they have desirable statistical properties as the probabilities are bound between 0 and 1 (Greene
2003).
Given that we investigate several adaptation choices, the appropriate econometric model would,
thus, be either a multinomial logit (MNL) or multinomial probit (MNP) regression model. Both models
estimate the effect of explanatory variables on a dependent variable involving multiple choices with
unordered response categories.
In this study, therefore, an MNL specification is adopted to model climate change adaptation
behavior of farmers involving discrete dependent variables with multiple choices. Thus, the probability
that household i with characteristics X chooses adaptation option j is specified as follows:
The main attractive feature of the MNP model is that it allows a rather general covariance
structure for the alternative-specific errors. However, because observed choices only reveal information
regarding utility differences, and because scale cannot be determined, not all parameters in an arbitrary
MNP specification may be identified (Bunch 1991). With recent advances in computational methods,
some researchers have developed techniques for the estimation of the MNP. A more recent method is the
Geweke-Hajivassiliou-Keane (GHK) smooth recursive conditioning stimulator used to calculate
multivariate normal probabilities of maximum likelihood estimation to evaluate trivariate and higher
dimensional normal distributions, based on the works of Cappellari and Jenkins (2003, 2005, 2006).
17
However, there is still a relative dearth of successful empirical applications of MNP in the published
literature (Bunch 1991; Tizale 2007).
Pij  prob(Y  1)

e x

j
1 e
, j=1….j,
(3)
x 
j 1
where β is a vector of parameters that satisfy ln (Pij/Pik) = X’(βj- βk) (Greene 2003).
Unbiased and consistent parameters estimates of the MNL model in Equation 3 require the
assumption of independence of irrelevant alternatives (IIA) to hold. Specifically, the IIA assumption
requires that the likelihood of a household’s using a certain adaptation measure needs to be independent
of other alternative adaptive measures used by the same household. Thus, the IIA assumption involves the
independence and homoscedastic disturbance terms of the adaptation model in Equation 1. The validity of
the IIA assumption could be tested using Hausman’s specification, which is based on the fact that if a
choice set is irrelevant, eliminating a choice or choice sets from the model altogether will not change
parameter estimates systematically.
Differentiating Equation 3 with respect to each explanatory variable provides marginal effects of
the explanatory variables given as
Pj
j 1


 Pj   jk   Pj  jk  .
xk
j 1


(4)
Using the MNL model for our analysis gives rise to a sample selectivity problem because only
those who perceive climate change will adapt. Indeed, adaptation to climate change begins with two
processes: first, perceiving change, and then, deciding on a particular adaptation choice. Therefore, the
correct modeling of the adaptation behavioral to climate change implies the use of a sample selectivity
model or a two-period framework (t=0, observation of the climate change; and t=1, adaptation choice is
made). However, the two-period choice model is a complicated model to estimate (Dagsvik 2000). Also,
a review of applied literature6 on selectivity bias correction methods did not reveal a two-stage model
that uses a multivariate choice model as the outcome equation. Therefore, both the Heckman sample
selectivity probit model and the MNL model are used to study the determinants of adaptation to climate
change.

Heckman sample selectivity probit model
Following Maddison (2006), Heckman’s sample selectivity probit model is based on the following two
latent variables:
Y1 = b'X + U1
(5)
Y2 = g'Z + U2,
where X is a k-vector of regressors; Z is an m-vector of regressors, possibly including 1's for the
intercepts; and the error terms U1 and U2 are jointly normally distributed, independently of X and Z, with
zero expectations. Although we are primarily interested in the first model, the latent variable Y1 is only
observed if Y2 > 0. Thus, the actual dependent variable is:
6
We found in the literature that following the seminal insight of Heckman (1979), two traditional approaches suggested by
Lee (1983) and Dubin and McFadden (1984) are applied when the selection model is a multinomial logit, and in the outcome
equation the dependent variable is a continuous variable. We also found that Dagsvik (2000) has developed a theoretical model
where the two stages are MNL model; however, a successful application of this model still has to be verified.
18
(6)
Y = Y1 if Y2 > 0, Y is a missing value if Y2 <= 0.
(7)
The latent variable Y2 itself is not observable, only its sign. We only know that Y2 > 0 if Y is
observable, and Y2 <= 0 if not. Consequently, we may without loss of generality normalize U2 such that
its variance is equal to 1. If we ignore the sample selection problem and regress Y on X using the
observed Y's only, then the ordinary least squares (OLS) estimator of b will be biased, because
E[Y1|Y2 > 0, X,Z] = b'X + rsf(g'Z)/F(g'Z),
(8)
where F is the cumulative distribution function of the standard normal distribution, f is the corresponding
density, s2 is the variance of U1, and r is the correlation between U1 and U2. Hence,
E[Y1|Y2 > 0, X] = b'X + rsE[f(g'Z)/F(g'Z)|X].
(9)
The latter term causes sample selection bias if r is nonzero. In order to avoid the sample selection
problem, and to get asymptotically efficient estimators, the model parameters are estimated by maximum
likelihood.
6.3. Choice of Variables and Hypotheses to be Tested
Based on the above information about adaptation choices, the choice sets considered in the adaptation
model include 10 variables:
1. Change crop variety
2. Intensification irrigation
3. Water-harvesting scheme
4. Plant different crops
5. Crop diversification/mixing
6. Change planting date
7. Change amount of land
8. Livestock feed supplements
9. Other
10. No adaptation
Much of what we know about the research question (farm and farmers’ characteristics related to
adaptive capacity and propensity) in the adaptation process derives from the vast body of research on the
dynamics of agricultural development and the diffusion of agricultural practices. Based on the review of
literature on adoption of new technologies and adaptation studies, a range of household and farm
characteristics, institutional factors, and other factors that describe local conditions are hypothesized to
influence farmers’ adaptation choice in the Limpopo River Basin.

Household characteristics
Generally the household characteristics considered to have differential impacts on adoption or adaptation
decisions are age, education level and gender of the head of the household, family size, years of faming
experience, and wealth.
According to Adesina and Forson (1995) cited by Teklewold et al. (2006), there is no agreement
in the adoption literature on the effect of age. The effect of age is generally location- or technologyspecific. The expected result of age is an empirical question. We may found that age negatively influence
the decision to adopt new technologies. It may be that older farmers are more risk-averse and less likely
to be flexible than younger farmers and thus have a lesser likelihood of adopting new technologies. In
another case, we may found that age positively influence the decision to adopt. It could also be that older
19
farmers have more experience in farming and are better able to assess the characteristics of modern
technology than younger farmers, and hence a higher probability of adopting the practice.
Higher level of education is often hypothesized to increase the probability of adopting new
technologies (Daberkow and McBride 2003; Adesina and Forson 1995). Indeed, education is expected to
increase one's ability to receive, decode, and understand information relevant to making innovative
decisions (Wozniak 1984).
Gender of the household head is hypothesized to influence the decision to adopt changes. The
way gender influences adaptation is location-specific. A number of studies in Africa have shown that
women have lesser access to critical resources (land, cash, and labor), which often undermines their
ability to carry out labor-intensive agricultural innovations (De Groote and Coulibaly 1998, Quisumbing
et al. 1995). However, a recent study by Nhemachena and Hassan (2007), based on Southern Africa, finds
that female-headed households are more likely to take up climate change adaptation methods. According
to the authors, the possible reason for this observation is that in most rural smallholder farming
communities in the region, men are more often based in towns, and much of the agricultural work is done
by women. Therefore, women have more farming experience and information on various management
practices and how to change them, based on available information on climatic conditions and other factors
such as markets and food needs of the households.
Wealth is believed to reflect past achievements of households and their ability to bear risks. Thus,
households with higher income and greater assets are in better position to adopt new farming technologies
(Shiferaw and Holden, 1998).
Farming experience increases the probability of uptake of all adaptation options because
experienced farmers have better knowledge and information on changes in climatic conditions and crop
and livestock management practices (Nhemachena and Hassan 2007).
The influence of household size on the decision to adapt is ambiguous. Household size as a proxy
to labor availability may influence the adoption of a new technology positively as its availability reduces
the labor constraints (Teklewold et al. 2006). However, according Tizale (2007), there is a possibility that
households with many family members may be forced to divert part of the labor force to off-farm
activities in an attempt to earn income to ease the consumption pressure imposed by a large family size.

Farm characteristics
Institutional factors often considered in the literature to influence adoption of new technologies are access
to information via extension services (climate information and production technologies), access to credit,
off-farm employment, and land tenure.
Agricultural extension enhances the efficiency of making adoption decisions. In the world of lessthan-perfect information, the introduction of new technologies creates a demand for information useful in
deciding on adopting new technologies (Wozniak 1984). Of the many sources of information available to
farmers, agricultural extension is the most important for analyzing the adoption decision. Based on the
innovation-diffusion literature (Adesina and Forson 1995), it is hypothesized that access to extension
services is positively related to adoption of new technologies by exposing farmers to new information and
technical skills. Also, in the specific case of climate change adaptation, access to climate information may
increase the likelihood of uptake of adaptation techniques.
Another variable that has received attention is access to credit, which commonly has a positive
effect on adaptation behavior (Caviglia-Harris 2002; Saín and Barreto 1996; Napier 1991; and Hansen et
al. 1987). Any fixed investment requires the use of owned or borrowed capital. Hence, the adoption of a
technology requires a large initial investment, which may be hampered by lack of borrowing capacity (El
Osta and Morehart 1999).
The occupation of the farmer is an indication of the total amount of time available for farming
activities. Off-farm employment may present a constraint to adoption of technology because it competes
for on-farm managerial time (McNamara et al. 1991).
Similarly, land tenure can contribute to adaptation, because landowners tend to adopt new
technologies more frequently than tenants, an argument that has justified numerous efforts to reduce
20
tenure insecurity (Lutz et al. 1994; Shultz et al. 1997). Land ownership is widely believed to encourage
the adoption of technologies linked to land such as irrigation equipment or drainage structures. Land
ownership is likely to influence adoption if the innovation requires investments tied to land.

Others factors
Local climatic conditions and agro-ecological conditions are expected to influence the decision to adapt.
We therefore included district level climate variables (temperature and rainfall). Further to take into
account spatial autocorrelation and neighborhood effects, we included latitude and longitude references
for each household. Finally we included dummy variables for provinces to take into account any specific
institutional arrangements having bearing on the ability of their farmers to adapt to climate change.
Table 11 provides the variables hypothesized to determine adaptation behavior, a brief
description of each variable, its value, and expected sign in relation to adoption of new technologies.
The farm characteristics hypothesized to influence adaptation in this study are farm size (largescale or small-scale) and soil fertility.
Farm size. Adoption of an innovation will tend to take place earlier on larger farms than on
smaller farms. Daberkow and McBride 2003 show that given the uncertainty and the fixed transaction and
information costs associated with innovation, there may be a critical lower limit on farm size that prevents
smaller farms from adapting. As these costs increase, the critical size also increases. It follows that
innovations with large fixed transaction and/or information costs are less likely to be adopted by smaller
farms.
Soil fertility. Farmers’ perceptions that their lands are infertile may be a first step in the
adaptation process. They may therefore be more likely to adopt any adaptation techniques that will help
improve their productivity.

Institutional factors
Institutional factors often considered in the literature to influence adoption of new technologies are access
to information via extension services (climate information and production technologies), access to credit,
off-farm employment, and land tenure.
Agricultural extension enhances the efficiency of making adoption decisions. In the world of lessthan-perfect information, the introduction of new technologies creates a demand for information useful in
deciding on adopting new technologies (Wozniak 1984). Of the many sources of information available to
farmers, agricultural extension is the most important for analyzing the adoption decision. Based on the
innovation-diffusion literature (Adesina and Forson 1995), it is hypothesized that access to extension
services is positively related to adoption of new technologies by exposing farmers to new information and
technical skills. Also, in the specific case of climate change adaptation, access to climate information may
increase the likelihood of uptake of adaptation techniques.
Another variable that has received attention is access to credit, which commonly has a positive
effect on adaptation behavior (Caviglia-Harris 2002; Saín and Barreto 1996; Napier 1991; and Hansen et
al. 1987). Any fixed investment requires the use of owned or borrowed capital. Hence, the adoption of a
technology requires a large initial investment, which may be hampered by lack of borrowing capacity (El
Osta and Morehart 1999).
The occupation of the farmer is an indication of the total amount of time available for farming
activities. Off-farm employment may present a constraint to adoption of technology because it competes
for on-farm managerial time (McNamara et al. 1991).
Similarly, land tenure can contribute to adaptation, because landowners tend to adopt new
technologies more frequently than tenants, an argument that has justified numerous efforts to reduce
tenure insecurity (Lutz et al. 1994; Shultz et al. 1997). Land ownership is widely believed to encourage
the adoption of technologies linked to land such as irrigation equipment or drainage structures. Land
ownership is likely to influence adoption if the innovation requires investments tied to land.
21

Others factors
Local climatic conditions and agro-ecological conditions are expected to influence the decision to adapt.
We therefore included district level climate variables (temperature and rainfall). Further to take into
account spatial autocorrelation and neighborhood effects, we included latitude and longitude references
for each household. Finally we included dummy variables for provinces to take into account any specific
institutional arrangements having bearing on the ability of their farmers to adapt to climate change.
Table 11 provides the variables hypothesized to determine adaptation behavior, a brief
description of each variable, its value, and expected sign in relation to adoption of new technologies.
Table 11. Variables hypothesized to affect adaptation decisions by farmers in the Limpopo River
Basin
Variable
Age
Education
Gender
Household size
Description
Household characteristics
Age of the head of the farm household
Value
Expected sign
Years
Number of years of formal schooling
attained by the head of the household
Gender of the head of the farm household
Years
Cannot be signed a
priori (+ or -)
Positive
Number of family members of a
household
Number of years of farming experience
for the household head
An index7 was constructed using
household ownership of seven
households’ assets: a television, radio,
flushing toilet, cell phone, brick house,
refrigerator, and car.
Farm characteristics
Determine if the farm is large-scale or
small-scale
Farmer’s own perception of the fertility
level of his land. Three dummies:
infertile, fertile, and highly fertile.
Farming experience
Wealth
Farm size
Soil fertility
1= male, 0=
female
Number
Years
Cannot be signed a
priori (+ or -)
Cannot be signed a
priori (+ or -)
Positive
number
Positive
1= large scale
0= small scale
0 or 1
Positive
Positive
7
Following Filmer and Pritchett 2001, principal component analysis (PCA) was used to assign weights to each asset. The
overall wealth index is calculated by applying the following formula:


w j   bi a ji  xi  / si
k
i 1
where w is the wealth index, b is the weights from PCA 1, a is the asset value, x is the mean asset value, and s is the standard
deviation of the assets.
22
Table 11. Continued
Variable
Extension
Climate information
Credit
Off-farm employment
Tenure
Temperature
Rainfall
Latitude
Description
Institutional factors
If household has access to extension
services
If household gets information about
weather, climate from any source
(extension officers, TV, radio, etc.)
If household has access to credit from any
sources
Income from off-farm activities during the
survey year
If land use is owned or rented/sharecropped, etc.
Others factors
Average temperature between 1960 and
2003
Average rainfall between 1960 and 2003
Value
1=yes, 0= no
Positive
1=yes, 0= no
Positive
1=yes, 0= no
Positive
1= owned
0= otherwise
Cannot be signed a
priori (+ or -)
Positive
degree Celsius
Positive
Negative
Limpopo
If household farm in Limpopo province
Mm
degree
centigrade
degree
centigrade
1=yes, 0= no
North west
If household farm in North West province
1=yes, 0= no
Gauteng
If household farm in Gauteng province
1=yes, 0= no
Mpumalanga
If household farm in Mpumalanga
province
1=yes, 0= no
Longitude
Expected sign
Cannot be signed a
priori (+ or -)
Cannot be signed a
priori (+ or -)
Cannot be signed a
priori (+ or -)
Cannot be signed a
priori (+ or -)
6.4. The Empirical Models and Results
In this section, two models for adaptation choices to climate change in the Limpopo River Basin are
estimated by using the statistical software Stata version 9.0: the Heckman probit and the MNL model. The
analysis is based on cross-sectional data collected in the Limpopo River Basin. Before initial runs of the
model, the data were checked for the presence of any multicollinearity in the data set. Among the
variables hypothesized to influence adaptation, age of the head of household was found to be correlated
inversely with education (ρ = -0.39) and positively with farming experience (ρ = 0.30) and extension and
climate information (ρ = 0.25). Although these correlations coefficients do not suggest incidence of
strong collinearity, age was dropped from the model (see Appendix Table B.4). Appendix Table B.5
provides the summary statistics of the independent variables included in the analysis.
6.4.1. Modeling Adaptation with the Heckman Probit Model
This section presents the results of the Heckman probit adaptation model. The model determines the
likelihood of perceiving any change in the climate as well as the likelihood of farmers’ adapting to these
changes. The dependent variable for the selection equation is binary indicating whether or not a farmer
perceives climate change; the dependent variable for the outcome equation is also binary indicating
whether or not a farmer responded to the perceived changes by adapting farming practices. The
explanatory variables are those discussed in the previous section, and “access to water for irrigation” is
also included in the selection equation. The likelihood function for the Heckman probit model was
significant (Wald χ2 = 36.26 with P<0.001), showing a strong explanatory power.
23
As shown in Table 12, access to water for irrigation, access to irrigation services, and living in
Gauteng influence the likelihood of perceiving climate change. On the other hand, farming experience,
farm size, soil fertility, access to extension, access to credit, land tenure status, and region influence the
probability of adapting to climate change.
Table 12. Results of the Heckman probit model of adaptations behavior in the Limpopo River
Basin
Variables
Access to water for irrigation
Education
Gender
Farming experience
Wealth
Farm size
High fertility of soil
Extension
Access to climate info
Credit
Off-farm employment
Land tenure
Mpumalanga
Gauteng
North West
Intercept
Wald test (zero slopes)
Wald test (independent
equations)
Total observations
Censored observations
Estimated coefficients
outcome equation: adaptation model
-0.011
0.134
0.01***
0.114
0.649***
-0.142*
0.179*
-0.1
0.232*
0.127
0.268***
-0.006
-0.603***
-0.445***
-0.6615***
36.26***
0.47
Estimated coefficients selection
equation: perception model
-0.621***
-0.012
-0.088
0.006
0.051
-0.036
-0.005
0.364***
-0.115
-0.0650
0.0472
0.0359
-0.031
-0.527**
-0.029
1.83***
577
43
Note: *** significant at 1% level ** significant at 5% level * significant at 10% level
As expected, experienced farmers are more likely to adapt. Likewise, farm size positively and
significantly leads to an increase in the likelihood of adapting to climate change. Farmers’ perception of
having highly fertile soil decreases the probability of taking up adaptation in response to changes in the
climate. Access to extension services increases the likelihood of perceiving changes in climate as well as
the likelihood of adaptation. This suggests that extension services help farmers to take climate changes
and weather patterns into account and help advise them on how to tackle climatic variability and change.
The results also show important regional variation. Farmers in Limpopo province are more likely to adapt
compared with farmers in the other provinces. Indeed, in Limpopo, the population is largely rural (82
percent), and the main rural economic activity is agriculture.
6.4.2. Modeling Adaptation with the Multinomial Logit Model
This section presents the empirical results of the MNL adaptation model. The MNL, as specified in
section 6.3 with 10 adaptation options, failed to produce satisfactory results in terms of the significance
level of the parameters estimates and also in terms of the validity of the independence of irrelevant
alternatives (IIA) assumption. The model was thus restructured by grouping three closely related choices
together in the same category. The replacement of plant-type cultivars with new varieties, the adoption of
new crops, and multicropping and mixed farming systems of crops and livestock were grouped in the
same category, labeled “portfolio diversification.” Indeed, these three choices are closely related because
they are considered for the same purpose of risk-spreading. Also, we aggregate intensification irrigation
24
and water-harvesting scheme into the same category, irrigation, because both are related to the use of
water for the purpose of increasing productivity and withstanding rainwater shortages.
Accordingly, the choice set in the restructured MNL model included the following adaptation
options:
1. Portfolio diversification
2. Irrigation
3. Change planting date
4. Change amount of land
5. Livestock feed supplements
6. Other
7. No adaptation
The MNL adaptation model with these restructuring choices was run and tested for the IIA
assumption using the Hausman specification test. The test failed to reject the null hypothesis of
independence of the included “change amount of land under cultivation,” suggesting there is no evidence
against the correct specification for the adaptation model, χ2 = -0.375 with P value of 0.9311. Therefore,
the application of the MNL specification to the data set for modeling climate change adaptation behavior
of farmers is justified.
Appendix Table B.6 and Table 13 present the estimated coefficients and the marginal effects,
respectively. The likelihood ratio statistics as indicated by χ2 = 291.07 are highly significant at 1 percent,
suggesting strong explanatory power of the model. It is important to note that the estimated coefficients
should be compared with the base category of not adopting any of the adaptation choices.
There is a 70 percent probability of farmers not adapting in face of climate change and variability.
Following are the household characteristics included in the model:
Household size. A large household will be more willing to choose the “Other” category as an
adaptation option. The “Other” category includes adaptations such as soil conservation techniques,
chemical treatments that are labor-intensive especially in small-scale farming, which involves household
labor.
Farming experience. Experienced farmers have an increased likelihood of using portfolio
diversification, changing planting dates, and changing the amount of land under production. These results
confirm the findings of Nhemachena and Hassan (2007) in a similar study of adaptation in the Southern
Africa region. Experienced farmers have high skills in farming techniques and management and are able
to spread risk when facing climate variability by exploiting strategic complementarities between activities
such as crop-livestock integration.
Wealth. Wealthier households are more willing to adapt by changing their planting dates.
Surprisingly, the results suggested that education level and gender did not have a significant impact on the
probability of choosing any adaptation technique.
The farm characteristics included in the model are as follows:



Farm size. The coefficient on farm size is significant and positively correlated with the
probability of choosing irrigation as an adaptation measure. Indeed, large-scale farmers are
more likely to adapt because they have more capital and resources. Therefore, they can easily
invest in irrigation technologies, which demand high investment costs.
Soil fertility. The perception of having highly fertile soil increases the probability that
farmers will change their amount of land under cultivation.
Following are the institutional factors:
Access to extension services. With an increased probability of taking up portfolio
diversification, farmers who have access to extension services are more likely to be aware of
changing climatic conditions (confirmed by the Heckman probit model, above) and to have
knowledge of the various management practices that they can use to adapt to changes in
25
climatic conditions. It appears that extension messages emphasized risk spreading and farmlevel risk management. Having access to extension increases the probability of choosing
portfolio diversification by 4 percent. The implementation of the land reform has increased
the number of new, emerging farmers who did not have the skills and information gathered
by experienced farmers; therefore, extension is of great need in South Africa.




Access to credit. As expected, access to credit increases the likelihood of adaptation. Poverty
or lack of financial resources is one of the main constraints to adjustment to climate change.
In a study on Tanzania, O’Brien et al. (2000) reports that despite numerous adaptation
options that farmers are aware of and willing to apply, the lack of sufficient financial
resources to purchase the necessary inputs and other associated equipment (e.g., purchasing
seeds, acquiring transportation, hiring temporary workers) is one of the significant constraints
to adaptation. In our study, 60 percent of the respondents who did not adapt cited lack of
financial resources as the main constraint to adaptation. The results show that access to credit
increases the likelihood that farmers will take up portfolio diversification and buy feed
supplements for their livestock. Having access to credit indeed increased the likelihood of
choosing portfolio diversification by 3 percent.
Tenure. Having secure property rights increases the probability of farmers to adapt by 9
percent. With proper property rights, farmers may be able change their amount of land under
cultivation to adjust to new climatic conditions.
Off-farm employment. While off-farm employment may present a constraint to adaptation
because it competes for on-farm managerial time (McNamara et al. 1991), the empirical
results suggest that off-farm activities increase the likelihood of buying feed supplements for
the livestock. This suggests that expanding smallholder farmers’ access to off-farm sources of
income increases the probability that they will invest in farming activities.
Temperature and rainfall. Households living in regions with high temperatures have an
increased likelihood of adapting. These households are more likely to choose the following
adaptation options: (1) portfolio diversification, such as by changing their types of crops (e.g.,
from maize to sorghum, a more heat-tolerant crop); (2) intensification irrigation; and (3)
changing their planting dates. A decrease in rainfall is likely to push farmers to delay their
planting dates.
26
Table 13. Results marginal effects of the MNL adaptation model, Limpopo River Basin
Education
Gender
Household
size
Farming
experience
Wealth
Farm size
Highly fertile
soil
Infertile soil
Extension
Climate
information
Credit
Off-farm
employment
Tenure
Latitude
Longitude
Rainfall
Temperature
Probability
Portfolio
diversification
Irrigation
-0.0023
(0.39)
-0.0084
(0.75)
-0.0021
(0.60)
0.0020
(0.01)***
-0.0083
(0.29)
0.0536
(0.32)
0.0342
(0.21)
-0.0375
(0.29)
0.0434
(0.09)*
-0.0257
(0.32)
0.0355
(0.06)*
0.0302
(0.27)
0.0112
(0.63)
0.0404
(0.03)**
-0.0132
(0.39)
0.0005
(0.12)
0.0133
(0.04)**
0.09
0.0019
(0.50)
0.0388
(0.22)
0.0058
(0.25)
0.0007
(0.59)
0.0128
(0.23)
0.1176
(0.09)*
0.0314
(0.39)
-0.0168
(0.73)
-0.0075
(0.80)
0.0018
(0.95)
0.0289
(0.42)
-0.0046
(0.88)
0.0204
(0.52)
-0.0208
(0.12)
-0.0020
(0.91)
0.0004
(0.16)
0.0136
(0.09)*
0.12
Changed
planting
dates
-0.0003
(0.82)
0.0115
(0.38)
-0.0002
(0.94)
0.0011
(0.03)**
0.0231
(0.00)***
0.0034
(0.91)
-0.0066
(0.64)
0.0091
(0.70)
0.0138
(0.30)
-0.0112
(0.43)
-0.0014
(0.93)
0.0006
(0.96)
0.0102
(0.47)
-0.0082
(0.10)*
0.0175
(0.02)**
-0.0003
(0.03)**
-0.0114
(0.00)***
0.035
Changed
amount
of land
0.0003
(0.56)
-0.0034
(0.54)
0.0003
(0.79)
0.0005
(0.09)*
0.0030
(0.22)
0.0077
(0.58)
0.0125
(0.10)*
0.0176
(0.30)
0.0052
(0.35)
0.0031
(0.60)
-0.0093
(0.19)
-0.0077
(0.09)*
0.0124
(0.10)*
0.0074
(0.14)
0.0060
(0.17)
0.0001
(0.36)
0.0005
(0.73)
0.01
Notes: *** significant at 1%, ** significant at 5%, * significant at 10%
27
Livestock
feed
supplements
-0.0003
(0.62)
0.0046
(0.41)
-0.0010
(0.25)
-0.0002
(0.47)
0.0010
(0.49)
-0.0007
(0.94)
0.0080
(0.32)
-0.0032
(0.64)
0.0016
(0.73)
-0.0011
(0.82)
0.0149
(0.09)*
0.0339
(0.00)***
-0.0048
(0.27)
-0.0034
(0.13)
-0.0039
(0.05)**
0.0000
(0.87)
0.0005
(0.72)
0.01
Other
No
Adaptation
0.0009
(0.49)
-0.0044
(0.8)
-0.0041
(0.09)*
-0.0001
(0.8)
0.0026
(0.62)
0.0030
(0.9)
-0.0148
(0.33)
0.0471
(0.20)
-0.0027
(0.84)
0.0172
(0.26)
0.0172
(0.37)
0.0074
(0.63)
0.0466
(0.02)**
0.0020
(0.69)
-0.0046
(0.52)
-0.0001
(0.68)
0.0075
(0.07)*
0.035
-0.0003
(0.94)
-0.0387
(0.37)
0.0013
(0.85)
-0.0039
(0.03)**
-0.0343
(0.01)***
-0.1846
(0.05)**
-0.0648
(0.17)
-0.0162
(0.81)
-0.0537
(0.08)*
0.0161
(0.69)
-0.0858
(0.08)*
-0.0597
(0.18)
-0.0960
(0.03)**
-0.0174
(0.41)
0.0002
(0.99)
-0.0006
(0.12)
-0.0240
(0.03)**
0.7
7. CONCLUSIONS AND POLICY IMPLICATIONS
The statistical analysis of temperature data from 1960 to 2003 in the Limpopo River Basin shows a trend
of increasing around 1 degree Celsius, with the increase mostly in the summer period. Over the 43 years
examined, rainfall is characterized by large interannual variability with a substantial decrease in the
amount of rainfall over the final three years of the data. However, there is a noticeable, long-running trend
of decreasing rainfall during winter.
Farmers’ perceptions of climatic variability are in line with climatic data records. Indeed, farmers
in the Limpopo River Basin of South Africa are able to recognize that temperatures have increased and
there has been a reduction in the volume of rainfall. Farmers with access to extension services are likely
to perceive changes in the climate because extension services provide information about climate and
weather. Having access to water for irrigation increases the resilience of farmers to climate variability;
therefore, they do not need to pay as much attention to changes in the patterns of rainfall and temperature.
With more experience, farmers are more likely to perceive change in temperature.
Although farmers are well aware of climatic changes, few seem to take steps to adjust their
farming activities. Only approximately 30 percent of farmers have adjusted their farming practices to
account for the impacts of climate change. The main adaptation strategies of farmers in the Limpopo
River Basin are switching crops, changing crop varieties, changing planting dates, increasing irrigation,
building water-harvesting schemes, changing the amount of land under cultivation, and buying livestock
feed supplements.
The Heckman probit and multinomial logit models are applied to examine the determinants of
adaptation to climate change and variability. The results highlight that household size, wealth, farm size,
farming experience, perception of soil fertility, extension, access to credit, off-farm activities, property
rights, high temperature, and low rainfall are the factors that enhance adaptive capacity to climate change.
Government policies should therefore ensure that farmers have access to affordable credit to
increase their ability and flexibility to change production strategies in response to the forecasted climate
conditions. Because access to water for irrigation increases the resilience of farmers to climate variability,
irrigation investment needs should be reconsidered to allow farmers increased water control to counteract
adverse impacts from climate variability and change. However, to promote efficient water use, emphasis
should be on pricing reforms and clearly defined property rights, as well as on the strengthening of farmlevel managerial capacity of efficient irrigation. More importantly, the implementation of the land reform
has increased the number of new, emerging farmers who did not have the skills and information gathered
by experienced farmers; therefore, increasing farmers’ access to extension services is of great need in
South Africa. Furthermore, government should improve off-farm income-earning opportunities.
28
APPENDIX A: QUESTIONNAIRE ON CLIMATE CHANGE AND ADAPTATION
OPTIONS
Section 9: Climate change and adaptation options
9.1 Have you noticed any long-term changes in the mean temperature over the last 20 years? (please
explain) Please mark with x if used.
If too difficult: Has the number of hot days stayed the same, increased, or declined over the last 20
years? (please explain)
9.2 Have you noticed any long-term changes in the mean rainfall over the last 20 years? (please
explain)
If too difficult: Has the number of rainfall days stayed the same, increased, or declined over the last
20 years? (please explain)
9.3 What adjustments in your farming have you made to these long-term shifts in temperature? Please
list below.
9.4 What adjustments in your farming have you made to these long-term shifts in rainfall? Please list
below.
9.5
Check the answers for 9.3 and 9.4 and then ask for the ones not yet listed there: What additional
measures would you consider in the future?
9.5.1 Why did you not
1. change crop variety
2. build a water-harvesting scheme
3. implement soil conservation techniques
4. buy insurance
5. put trees for shading
6. irrigate more
7. change from crop to livestock
8. reduce number of livestock
9 migrate to urban area
10.find off-farm job
11.lease your land
9.5.2 Reason (key)
Key for 9.5.2: 1: lack of money, 2: lack of information, 3: shortage of labor, 5: Other [write into the lines
provided above]
9.6
What were the main constraints/difficulties in changing your farming ways?
9.7
Have you seen any changes in the number of malaria cases over the last 20 years? (please
explain)
9.8
Have you seen changes in the vegetation cover over the last 5 years? (please explain)
29
APPENDIX B: SUPPLEMENTARY TABLES
Table B.1. Perceptions of changes in temperature by farmer experience (%)
Total population
Low experience
0–10 years
Medium experience
10–30 years
High experience
30 years +
Increased
Decreased
Others
1.56
More or less
extreme
1.56
89.32
No change
2.21
5.08
Don't
know
0.26
89.01
1.76
1.32
2.64
5.05
0.22
87.50
1.39
2.78
1.85
6.02
0.46
94.85
1.03
0
1.03
3.09
0
Notes: Test: Equality of populations: The Kruskal-Wallis Test; chi-squared = 1.105 with 2 d.f.; probability = 0.5755
Table B.2. Perceptions of changes in rainfall by farmer experience (%)
Total
population
Low
experience
0– 0 years
Medium
experience
10– 0 years
High
experience
30 years +
Increased
Decreased
Change in
timing of
rains (earlier/
later/erratic)
Change in
frequency
of
droughts/
floods
Others
No
change
Don't
know
Decrease
in rainfall
and
change in
timing
1.69
81.25
5.34
0.65
1.43
2.21
0.39
7.03
1.10
78.51
6.14
0.66
1.75
2.41
0.44
8.99
2.79
86.05
4.19
0.47
0.93
1.86
0.47
3.26
2.06
83.51
4.12
1.03
1.03
2.06
0.00
6.19
Notes: Test: Equality of populations: The Kruskal-Wallis Test; chi-squared = 1.105 with 2 d.f.; probability = 0.5755
Table B.3. Perceptions of changes in temperature by farmer education level (%)
Total population
No formal education
0 years
Standard education
1–7 years
Secondary education
8–12 years
Tertiary education
13 years +
Increased
Decreased
Others
1.56
More or less
extreme
1.56
Don't know
2.21
No
change
5.08
89.32
88.00
2.00
2.67
2.00
4.67
0.67
91.06
1.63
0.00
0.81
6.50
0.00
89.06
1.29
1.80
2.58
4.38
0.26
87.85
1.87
0.93
2.80
6.54
0.00
0.26
Notes: Test: Equality of populations: The Kruskal-Wallis Test; chi-squared = 0.253 with 3 d.f.; probability = 0.9687
30
Table B.4. Correlation matrix of the independent variables of the adaptation model
Age
Education
Gender
Household
size
Experience
Wealth
Farm size
Infertile
soil
Fertile
soil
Highly
fertile
soil
Extension
Climate
information
Credit
Off-farm
activities
Tenure
Temperature
Rainfall
Latitude
Longitude
Limpopo
Gauteng
Age
Education
Gender
1
-0.3924
0.1329
1
0.0743
1
0.077
0.3055
-0.0774
-0.0243
-0.2063
0.0077
-0.0032
0.1391
0.0261
-0.0125
-0.0065
0.11
-0.0142
-0.0373
-0.0254
Household
size
Experience
Wealth
1
0.0093
-0.0478
-0.0035
1
-0.0189
0.057
1
0.0381
0.0212
0.0329
0.0381
0.0763
-0.062
-0.0666
0.0052
0.0274
-0.0452
0.0415
0.0921
0.0582
-0.0088
-0.0346
-0.0135
-0.001
-0.0165
-0.0493
-0.0233
-0.122
0.0578
0.083
-0.0625
-0.0751
0.117
0.0141
0.0347
0.042
-0.0081
0.0502
-0.0339
-0.0045
-0.0186
-0.1075
0.0208
-0.0588
0.1886
0.064
0.0269
0.052
0.1399
-0.0592
-0.0053
-0.0455
0.0887
-0.0583
-0.0977
0.0499
0.0055
0.0258
0.0713
0.0079
-0.0755
-0.0709
-0.024
-0.078
0.1206
-0.0797
0.1145
-0.0304
-0.0867
-0.1112
-0.0467
-0.0382
0.0788
0.146
0.0795
-0.1537
0.1868
-0.0421
0.1645
-0.025
-0.0102
0.0007
-0.0135
-0.0946
0.054
0.1038
-0.0421
-0.0909
-0.0586
0.0498
0.0396
-0.2081
0.0645
-0.0337
0.0387
0.0388
-0.1405
-0.1657
0.0392
0.0388
0.012
Mpumalanga
North West
31
Farm
size
Infertile
soil
1
0.0129
0.0561
1
0.0772
0.0115
0.0173
0.1037
0.0263
0.1246
0.0785
0.0172
0.0713
0.0039
0.0268
0.0475
0.0388
0.0461
Fertile
soil
Highly
fertile
soil
Extension
Climate
information
-0.4209
1
-0.2099
-0.7724
1
-0.0785
0.0114
0.027
1
-0.0595
0.0005
0.0694
-0.0052
-0.0448
-0.0162
0.2513
-0.0104
1
-0.1004
0.0685
-0.0715
0.0818
-0.0339
-0.0052
-0.1372
-0.0081
0.0765
0.0946
0.0309
0.0805
-0.0466
-0.0981
-0.0427
-0.0307
-0.1241
-0.0428
-0.0496
0.0188
0.1844
-0.1509
0.0609
-0.1906
-0.0937
0.1828
-0.0245
0.1791
0.1432
-0.0047
0.1268
-0.0518
-0.0083
0.0771
0.0519
-0.0277
-0.0202
0.0267
-0.0087
-0.0099
-0.0825
0.0234
-0.0446
0.2366
-0.047
-0.2584
0.08
0.1398
-0.1006
0.0494
0.0367
Table B.4. Continued
Credit
Age
Education
Gender
Household
size
Experience
Wealth
Farm size
Infertile
soil
Fertile
soil
Highly
fertile
soil
Extension
Climate
information
Credit
Off-farm
activities
Tenure
Temperature
Rainfall
Latitude
Longitude
Limpopo
Gauteng
Mpumalanga
North West
Off-farm
activities
Tenure
Temperature
Rainfall
Latitude
Longitude
1
0.05
-0.02
-0.05
1
-0.1987
0.1206
1
-0.3619
1
0.02
-0.08
-0.2303
-0.1096
0.3527
0.2796
-0.1676
0.5386
1
0.4346
1
0.03
0.07
-0.1
0.03
-0.2738
0.176
0.1598
0.0707
0.5573
-0.4985
-0.167
-0.251
-0.5423
0.0856
0.5845
-0.0228
0.7027
-0.1983
-0.5118
-0.2328
0.1218
-0.1576
0.3538
-0.5381
Limpopo
Gauteng
Mpumalanga
1
-0.2882
-0.6592
-0.4188
1
-0.1536
-0.0976
1
-0.2232
North
West
1
0.0848
0.0874
0.0011
0.0764
0.0672
0.0015
0.1053
0.0562
0.033
0.0735
32
1
Table B.5. Summary statistics of the variables for the adaptation model
Variables
Observations
Mean
Age
Education
Household size
Farming experience
771
775
784
784
55.32
7.39
6.22
12.51
Wealth
Latitude
Longitude
Temperature
Rainfall
781
794
794
794
794
Choices
0
1
Total
0
1
Total
0
1
Total
0
1
Total
0
1
Total
0
1
Total
0
1
Total
0
1
Total
0
1
Total
0
1
Total
0
1
Total
0
1
Total
0
1
Total
0.00000653
-24.58
29.02
20.45
564.45
Frequency
216
568
784
730
64
794
521
231
752
309
443
752
674
78
752
201
549
750
303
453
756
455
262
717
616
178
794
516
227
743
419
327
746
342
432
794
684
110
794
Gender
Farm size
Infertile soil
Fertile soil
Highly fertile soil
Access to water for
irrigation
Extension
Climate info
Credit
Off-farm
employment
Tenure
Limpopo
North West
Standard
Deviation
12.87
5.10
2.99
11.72
Minimum
Maximum
18
0
1
1
100
33
24
60
1.50
1.64
1.66
2.39
94.61
Percent
27.55
72.45
-3.96
-31.57
21.20
15.04
260.092
Cumulative
27.55
100
2.12
-22.33
31.75
23.03
814.184
91.94
8.06
91.94
100.00
69.38
30.72
69.28
100
41.09
58.91
41.09
100
89.63
10.37
89.63
100
26.80
73.20
100
40.08
59.92
100
63.46
36.54
100
77.58
22.42
100
69.45
30.55
100
56.17
43.83
100
43.72
54.48
26.80
100
86.13
13.87
33
39.66
100
63.46
100
77.58
100
69.45
100
56.17
100
43.72
100
86.13
100
Table B.5. Continued
Variables
Observations
Mean
Mpumalanga
0
1
Total
0
1
Total
0
1
Total
0
1
Total
0
1
Total
0
1
Total
0
1
Total
Notice temperature
change
Notice rainfall
change
Notice climate
change
Adapt to
temperature change
Adapt to rainfall
change
No Adaptation
Minimum
592
202
Standard
Deviation
74.55
25.45
41
727
768
20
748
768
54
711
765
494
223
717
497
241
738
298
463
761
5.34
94.66
100
2.60
97.40
100
7.06
92.94
100
68.50
31.10
100
67.34
32.66
100
39.16
60.84
100
5.34
100
34
74.55
100
2.60
100
7.06
100
68.50
100
67.34
100
60.84
100
Maximum
Table B.6. Results of the multinomial logit adaptation model, Limpopo River Basin
Portfolio
diversification
Irrigation
Changed
planting dates
Changed the
amount of land
Livestock feed
supplements
Other
Education
-0.0256
(0.45)
0.015
(0.56)
-0.0072
(0.84)
0.0239
(0.54)
-0.02606
(0.63)
0.0274
(0.51)
Gender
-0.0381
(0.90)
0.390
(0.24)
0.4163
(0.39)
-0.2365
(0.61)
0.5869
(0.41)
-0.0687
(0.89)
Household
size
-0.0256
(0.62)
0.045
(0.34)
-0.0065
(0.92)
0.0235
(0.80)
-0.1032
(0.23)
-0.1206
(0.09)*
Farming
experience
0.0283
(0.01)***
0.011
(0.36)
0.0362
(0.01)***
0.0478
(0.00)***
-0.0159
(0.55)
0.0015
(0.92)
Wealth
-0.0439
(0.66)
0.152
(0.09)*
0.7200
(0.00)***
0.3255
(0.07)*
0.1524
(0.32)
0.1223
(0.43)
Farm size
0.7903
(0.09)*
0.994
(0.03)**
0.3967
(0.65)
0.8507
(0.26)
0.2231
(0.86)
0.3850
(0.58)
Highly
fertile soil
0.4521
(0.10)*
0.335
(0.29)
-0.1061
(0.823)
1.0011
(0.05)**
0.7963
(0.19)
-0.3737
(0.50)
Infertile soil
-0.4928
(0.44)
-0.120
(0.81)
0.2631
(0.658)
1.0477
(0.14)
-0.3601
(0.71)
0.9341
(0.09)*
Extension
0.5876
(0.09)*
0.016
(0.95)
0.4915
(0.27)
0.5658
(0.22)
0.2494
(0.62)
0.0001
(0.99)
Climate
information
-0.3225
(0.33)
-0.009
(0.97)
-0.3602
(0.42)
0.2471
(0.61)
-0.1404
(0.80)
0.4426
(0.29)
Credit
0.4903
(0.08)*
0.348
(0.26)
0.0860
(0.87)
-0.9608
(0.13)
1.2228
(0.02)**
0.5648
(0.22)
Off-farm
employmen
t
0.4091
(0.19)
0.050
(0.87)
0.1044
(0.80)
-0.7191
(0.31)
2.1167
(0.00)***
0.2936
(0.50)
Tenure
0.2672
(0.36)
0.304
(0.31)
0.4309
(0.28)
1.1490
(0.02)**
-0.3775
(0.49)
1.3012
(0.00)***
Latitude
0.4806
(0.05)**
-0.142
(0.25)
-0.2132
(0.17)
0.7003
(0.09)*
-0.3234
(0.07)*
0.0809
(0.58)
35
Table B.6. Continued
Portfolio
diversification
Irrigation
Changed
planting dates
Changed the
amount of land
Livestock feed
supplements
Other
Longitude
-0.1496
(0.45)
-0.016
(0.92)
0.5093
(0.03)**
0.5458
(0.23)
-0.4082
(0.09)*
-0.1312
(0.54)
Rainfall
0.0062
(0.12)
0.004
(0.13)
-0.0071
(0.04)**
0.0079
(0.269)
0.0001
(0.97)
-0.0007
(0.86)
0.0855
(0.60)
0.2502
(0.03)**
-2.6090
(0.75)
-2.3139
(0.74)
Temperatur
e
0.1851
(0.02)**
6.1086
(0.49)
Intercept
Base
category
No.
observations
LR chisquare (90)
Log pseudo
likelihood
Pseudo RSquare
0.144
(0.08)*
-0.2957
(0.00)***
0.0783
(0.53)
-11.102
(0.05)**
-14.0650
(0.09)*
-10.6187
(0.53)
No Adaptation
591
291.07***
-676.2506
0.1320
Notes: *** significant at 1% probability level, ** significant at 5% probability level, * significant at 10% probability level
Hausman Test for the IIA Assumption
Test: Ho: difference in coefficients not systematic
chi2(17) = (b-B)'[(V_b-V_B)^(-1)](b-B)
= -0.375
Prob>chi2 =
0.9311
(V_b-V_B is not positive definite)
Fail to reject the null hypothesis; therefore, the MNL adaptation model holds consistent.
36
REFERENCES
Adesina, A.A, and J.B. Forson. 1995. Farmers' perceptions and adoption of new agricultural technology: Evidence
from analysis in Burkina Faso and Guinea, West Africa. Agricultural Economics 13:1–9.
Belliveau S., B. Bradshaw, B. Smit, S. Reid, D. Ramsey, M. Tarleton, and B. Sawyer. 2006. Farm-level adaptation
to multiple risks: Climate change and other concerns. Occasional paper No. 27. Canada: University of
Guelph.
Brklacich M., D. McNabb, C. Bryant, and I. Dumanski. 1997. Adaptability of agriculture systems to global climate
change: A Renfrew County, Ontario, Canada pilot study. In Agricultural restructuring and sustainability: A
geographical perspective, eds. B. Iibery, Q. Chiotti, and T. Richard. Wallingford: CAB International.
Bryant, R.C., B. Smit, M. Brklacich, R.T. Johnston, J. Smithers, Q. Chiotti, and B. Singh. 2000. Adaptation in
Canadian agriculture to climatic variability and change. Climatic Change 45:181–201.
Bunch, D.S. 1991. Estimability in the multinomial probit model. Transportation Research 25(1): 1–12.
Cappellari L. and S.P. Jenkins. 2003. Multivariate probit regression using simulated maximum likelihood. The Stata
Journal 3: 278–294.
Cappellari L. and S.P. Jenkins. 2005. Software update: st0041-1 Multivariate probit regression using simulated
likelihood estimation. The Stata Journal 5: 285.
Cappellari L. and S.P. Jenkins. 2006. Calculation of multivariate probabilities by simulation, with maximum
simulated likelihood estimation. The Stata Journal 6: 156–189.
Caviglia-Harris, J. 2002. Sustainable Agricultural Practices in Rondônia, Brazil: Do Local Farmer Organizations
Impact Adoption Rates? Department of Economics and Finance, Salisbury University.
Cliff, A. D., Ord, J. K. 1981. Spatial processes, Models and Applications. Pion, London.
Chiotti, Q., T.R.R. Johnston, B. Smit, and B. Ebel. 1997. Agricultural response to climate change: A preliminary
investigation of farm-level adaptation in Southern Alberta. In Agricultural restructuring and sustainability:
A geographical perspective, eds. B. Iibery, Q. Chiotti, and T. Richard. Wallingford: CAB International.
Daberkow, S.G. and W.D. McBride. 2003. Farm and Operator Characteristics Affecting the Awareness and
Adoption of Precision Agriculture Technologies in the U.S. Precision Agriculture. 4:163–177.
Dagsvik, J.K. 2000. Multinomial Choice and Selectivity. Discussion Papers 264, Research Department of Statistics
Norway.
De Groote, H. and N.Coulibaly. 1998. Gender and Generation: An Intra-Household Analysis on Access to
Resources in Southern Mali. African Crop Science Journal 6(1): 79–95.
Deressa, T.T. 2003. Measuring the impact of climate change on South African agriculture: The case of sugarcane
growing regions. MSc. Thesis, Faculty of Natural and Agricultural Sciences, University of Pretoria;
Pretoria, South Africa.
Du Toit, A.S., S. Prinsloo, and A. Marthinus. 2001. El Nino-Southern Oscillation effects on maize production in
South Africa. A preliminary methodology study. In Impacts of El Niño and climate variability on
agriculture, eds. C. Rosenzweig, K.J. Boote, S. Hollinger, A. Iglesias, and J.G. Phillips, 77–86. ASA
Special Publication 63, American Society of Agronomy; Madison, Wisc., USA.
DWAF (Department of Water Affairs and Forestry, South Africa). 2004 a. Internal Strategic Perspective: Limpopo
Water Management Area: Prepared by Goba Moahloli Keeve Steyn (Pty) Ltd, in association with Tlou &
Matji (Pty) Ltd and Golder Associates (Pty) Ltd. on behalf of the Directorate: National Water Resource
Planning. Report No. P WMA 01/000/00/0304, Pretoria, South Africa
Easterling, W.E., P.R. Crosson, N.J Rosenberg, M.S. McKenney, L.A. Katz, and K.M. Lemon. 1993. Agricultural
impacts of and responses to climate change in the Missouri-Iowa-Nebraska region. Climatic Change, 24
(1–2): 23–62.
El-Osta, H. and M. Morehart. 1999. Technology adoption decisions in diary production and the role of herd
expansion. Agricultural and Resource Economics Review 28(1):84–95.
Erasmus, B., A. van Jaarsveld, J. van Zyl, and N. Vink. 2000. The effects of climate change on the farm sector in
Western Cape. Agrekon 39 (4): 559–573.
Gbetibouo, G.A., and R.M. Hassan. 2005. Measuring the economic impact of climate change on major South
African field crops: a Ricardian approach, Global and Planetary Change 47 (2005) 143–152.
Greene, W. H. 2003. Econometric analysis, 5th ed. Prentice Hall, Upper Saddle River, New Jersey: Prentice-Hall.
Hageback, J., J. Sundberg, M. Ostwald, D. Chen, X. Yun and P. Knutsson. 2005. Climate variability and land use
change in Danagou, watershed, China—Examples of small scale farmers’ adaptation. Climatic Change 72:
189–212.
Hansen, D., J. Erbaugh, and T. Napier. 1987. Factors Related to Adoption of Soil Conservation Practices in the
Dominican Republic. Journal of Soil and Water Conservation 42: 367–369.
Heckman, J.J. 1979. Sample selection bias as a specification error. Econometrica 47 (1): 153–161.
IPCC, 2001. Climate Change 2001: Impacts, Adaptation, and Vulnerability. Report edited by McCarthy J.J. et al.,
Contribution of Working Group II to the Third Assessment Report of the Intergovernmental Panel on
Climate Change. Cambridge University Press, Cambridge, UK.
Kiker, G.A. 2002. CANEGRO-DSSAT linkages with geographic information systems: Applications in climate
change research for South Africa. Proceedings of International CANGRO Workshop; Mount Edgecombe,
South Africa.
Kiker, G.A., I.N. Bamber, G. Hoogenboom, M. Mcgelinchey. 2002. Further progress in the validation of the
CANEGRO-DSSAT model. Proceedings of International CANGRO Workshop; Mount Edgecombe, South
Africa.
Kolstad, C., D. Kelly, and G. Mitchell. 1999. Adjustment costs from environmental change induced by incomplete
information and learning. Working Paper 10-99. Department of Economics, University of California, Santa
Barbara.
Kurukulasuriya, P., and S. Rosenthal. 2003. Climate change and agriculture: A review of impacts and adaptations.
Climate Change Series paper no. 91. Environment Department and Agriculture and Rural Development
Department, World Bank, Washington, DC.
Lee L.F. 1983. Generalized econometric models with selectivity. Econometrica 51:507–512.
Lutz, E., S. Pagiola, y C. Reiche. 1994. The Costs and Benefits of Soil Conservation: The Farmer's Viewpoint. The
World Bank Research Observer 9: 273–295.
Maddison, D. 2006. The perception of and adaptation to climate change in Africa. CEEPA Discussion Paper No.
10. Centre for Environmental Economics and Policy in Africa, University of Pretoria, South Africa.
McFadden D.L. 1973. Conditional logit analysis of qualitative choice behavior. In Frontiers in econometrics, ed. P.
Zarembka. Academic Press, New York.
McNamara, K.T, M.E. Wetzstein and G.K. Douce. 1991. Factors affecting peanut producer adoption of integrated
pest management. Review of agricultural economics 13: 129–139.
Mendelsohn, R. 1998. Climate-change damages. In Economics and policy issues in climate change, ed. W.D.
Nordhaus. Resources for the Future: Washington, D.C.
Mendelssohn, R. 2001. Adaptation. In Global warming and the American economy: A regional assessment of
climate impacts, ed. R. Mendelsohn. Edward Elgar: Cheltenham: UK.
Napier, T.1991. Factors Affecting Acceptance and Continued Use of Soil Conservation Practices in Developing
Societies: A Diffusion Perspective. Agriculture, Ecosystems and Environment 36: 127–140.
Nhemachena, C., and R. Hassan. 2007. Micro-level analysis of farmers’ adaptation to climate change in Southern
Africa. IFPRI Discussion Paper No. 00714. International Food Policy Research Institute, Washington, D.C.
O’Brien, K., L. Sygna, L.O. Naess, R. Kingamkono and B. Hochobeb. 2000. Is information enough? User responses
to seasonal climate forecasts in Southern Africa. Report 2000:3, Center for International climate and
Environmental Research (CICERO), Oslo.
Poonyth, D., R.M. Hassan, G.A. Gbetibouo, J.M. Ramaila, and M.A. Letsoalo. 2002. Measuring the impact of
climate change on South African agriculture: A Ricardian approach. A paper presented at the “40th Annual
Agricultural Economics Association of South Africa Conference,” 18–20 September, Bloemfontein, South
Africa.
Quisumbing, A., L. Haddad and C. Peña. Gender and Poverty: New Evidence from 10 Developing Countries.
FCND Discussion Paper No. 9, International Food Policy Research Institute, Washington, D.C.
Reilly, J., and D. Schimmelpfennig. 1999. Agricultural impact assessment, vulnerability and the scope for
adaptation. Climatic change 43: 745–788.
Risbey J., M. Kandlikar, and H. Dowlatabadi. 1999. Scale, context and decision making in agricultural adaptation to
climate variability and change. Mitigation and Adaptation Strategies for Global Change 4 (2): 137–167.
Rosenzweig, C., and M.L. Parry. 1994. Potential impact of climate-change on world food supply. Nature 367:133–
138.
Schulze, R. 2003. Climate change and water resources: How vulnerable are we? What adaptations options do we
pursue? Paper presented at the training workshop on quality control for country-level and regional analyses
and reporting of the GEF/WB regional climate, water, and agriculture: Impacts on and adaptation of agroecological systems in Africa project; Cairo, Egypt.
Saín, G., and H. J. Barreto. 1996. The Adoption of Soil Conservation Technology in El Salvador: Linking
Productivity and Conservation. Journal of Soil and Water Conservation 51: 313–321.
Schulze, R.E., G.A. Kiker, and R.P. Kunz. 1993. Global climate change and agricultural productivity in Southern
Africa. Global Environ. Change 3 (4): 330-349.
Shiferaw, B. and S. Holden. 1998. Resource degradation and adoption of land conservation technologies in the
Ethiopian Highlands: A case study in Andit Tid, North Shewa, Agricultural Economics, 18:233–247
Shultz, S., J. Faustino and D. Melgar. 1997. Agroforestry and Soil Conservation: Adoption and Profitability in El
Salvador. Agroforestry Today 9: 16–17.
Smit B., and M.W. Skinner. 2002. Adaptations options in agriculture to climate change: A typology. Mitigation and
Adaptation Strategies for Global Change 7: 85–114.
Smit, B., D. McNabb, and J. Smithers. 1996. Agricultural adaptation to climatic variation. Climatic Change 33: 7–
29.
Smith, J.B. 1996. Using a decision matrix to assess climate change adaptation. In Adapting to climate change: An
international perspective, ed. J.B. Smith, N. Bhatti, G. Menzhulin, R. Benioff, M.I. Budyko, M. Campos,
B. Jallow, and F. Rijsberman. New York: Springer.
Smith J.B., and S. Lenhart. 1996. Climate change adaptation policy options. In Vulnerability and adaptation of
African ecosystems to global climate change, CR special, 6(2), book version.
Teklewold, H., L. Dadi, A. Yami, and N. Dana. 2006. Determinants of adoption of poultry technology: A doublehurdle approach. Livestock Research for Rural Development 18 (3).
http://www.lrrd.org/lrrd18/3/tekl18040.htm
Tizale, C.Y. 2007. The dynamics of soil degradation and incentives for optimal management in the Central
Highlands of Ethiopia. PhD thesis. Faculty of Natural and Agricultural Sciences, University of Pretoria;
Pretoria, South Africa.
Vedwan N. and R.E. Rhoades. 2001. Climate change in the Western Himalayas of India: a study of local perception
and response. Climate Research. 19: 109–117.
Wall, E., and B. Smit. 2005. Climate change adaptation in light of sustainable agriculture. Journal of Sustainable
Agriculture 27 (1): 113–123.
Wozniak, G.D. 1984. The adoption of interrelated innovations: A human capital approach. Review of Economics and
Statistics 66 (LXVI): 70–79.
RECENT IFPRI DISCUSSION PAPERS
For earlier discussion papers, please go to www.ifpri.org/pubs/pubs.htm#dp.
All discussion papers can be downloaded free of charge.
848.
Agglomeration, migration, and regional growth: A CGE analysis for Uganda. Paul Dorosh and James Thurlow, 2009.
847.
Biosafety decisions and perceived commercial risks: The role of GM-free private standards. Guillaume Gruère and
Debdatta Sengupta, 2009.
846.
Impact of soaring food price in Ethiopia: does location matter? John M. Ulimwenju, Sindu Workneh, and Zelekawork
Paulos, 2009.
845.
Aggregate effects of imperfect tax enforcement. Miguel Robles, 2009.
844.
Agricultural strategy development in West Africa: The false promise of participation? Danielle Resnick and Regina
Birner, 2008.
843.
Climate variability and maize yield in South Africa: Results from GME and MELE methods. Wisdom Akpalu, Rashid M.
Hassan, and Claudia Ringler, 2008.
842.
Local impacts of a global crisis: Food price transmission and poverty impacts in Ghana. Godsway Cudjoe, Clemens
Breisinger, and Xinshen Diao, 2008.
841.
Technology transfer, policies, and the role of the private sector in the global poultry revolution. Clare A. Narrod, Carl E.
Pray, and Marites Tiongco, 2008.
840.
The impact of agricultural extension and roads on poverty and consumption growth in fifteen Ethiopian villages. Stefan
Dercon, Daniel O. Gilligan, John Hoddinott, and Tassew Woldehanna, 2008.
839.
The impact of Ethiopia’s Productive Safety Net Programme and its linkages. Daniel O. Gilligan, John Hoddinott, and
Alemayehu Seyoum Taffesse, 2008.
838.
Aid effectiveness and capacity development: Implications for economic growth in developing countries. Prabuddha Sanyal
and Suresh Babu, 2008.
837.
A two-dimensional measure of polarization. Tewodaj Mogues, 2008.
836.
Higher fuel and food prices: Economic impacts and responses for Mozambique. Channing Arndt, Rui Benfica, Nelson
Maximiano, Antonio M.D. Nucifora, and James T. Thurlow, 2008
835.
Accelerating innovation with prize rewards: History and typology of technology prizes and a new contest design for
innovation in African agriculture. William A. Masters and Benoit Delbecq, 2008.
834.
Local politics, political institutions, and public resource allocation. Nethra Palaniswamy and Nandini Krishnan, 2008.
833.
Trade protection and tax evasion: Evidence from Kenya, Mauritius, and Nigeria. Antoine Bouet and Devesh Roy, 2008.
832.
Global carbon markets: Are there opportunities for Sub-Saharan Africa? Elizabeth Bryan, Wisdom Akpalu, Mahmud
Yesuf, and Claudia Ringler, 2008.
831.
Anatomy of a crisis: The causes and consequences of surging food prices. Derek Heady and Shenggen Fan, 2008
830.
Credit constraints, organizational choice, and returns to capital: Evidence from a rural industrial cluster in China.
Jianqing Ruan and Xiaobo Zhang, 2008.
829.
The future of global sugar markets: Policies, reforms, and impact. Proceedings of a public conference. David Orden,
Jean-Christophe Bureau, Alexandre Gohin, Loïc Guindé, Guy Millet, Antônio Salazar P. Brandão, Stephen Haley, Owen
Wagner, Ron Sandrey and Nick Vink, 2008.
828.
The impact of climate change and adaptation on food production in low-income countries: Evidence from the Nile Basin,
Ethiopia. Mahmud Yesuf, Salvatore Di Falco, Claudia Ringler, and Gunnar Kohlin, 2008.
827.
The Philippines: Shadow WTO agricultural domestic support notifications. Caesar Cororaton, 2008.
826.
What determines adult cognitive skills? Impacts of preschooling, schooling, and post-schooling experiences in
Guatemala. Jere R. Behrman, John Hoddinott, John A. Maluccio, Erica Soler-Hampejsek, Emily L. Behrman, Reynaldo
Martorell, Manuel Ramírez-Zea, and Aryeh D. Stein, 2008.
INTERNATIONAL FOOD POLICY
RESEARCH INSTITUTE
www.ifpri.org
IFPRI HEADQUARTERS
2033 K Street, NW
Washington, DC 20006-1002 USA
Tel.: +1-202-862-5600
Fax: +1-202-467-4439
Email: [email protected]
IFPRI ADDIS ABABA
P. O. Box 5689
Addis Ababa, Ethiopia
Tel.: +251 11 6463215
Fax: +251 11 6462927
Email: [email protected]
IFPRI NEW DELHI
CG Block, NASC Complex, PUSA
New Delhi 110-012 India
Tel.: 91 11 2584-6565
Fax: 91 11 2584-8008 / 2584-6572
Email: [email protected]